Version History

Expand All

Version 8 History

02 November 2015

The following bug fixes have been made in SSI Web in general:

  1. Additional security measures have been implemented.
  2. In rare circumstances in the export area an error "startindex cannot be larger than the length of the string" would occur.
  3. An unnessary warning would occur when exporting the field sys_SequentialRespNum.
  4. Auto upload would display the error "unsupported encryption" on FIPS 140-2 compliant computers.
  5. If a filter used by an export job is deleted and the export job is run, an error would occur.
  6. If a quota question came first on a page with multiple questions, then backing up would cause the original quota value to be removed when the page was resubmitted.

The following bug fixes have been made in ACBC:

  1. ACBC Interaction Search Tool with optional .hbu file for Main Effects would show "Matrices are not compatible" error in some cases.
Read More

04 August 2015

The following bug fixes have been made in SSI Web in general:

  1. Web servers running Apache Mod_Perl could not run surveys. Either a Sawtooth Software error #313 would be displayed or a generic server error would occur. To completely fix this problem on your Mod_Perl enabled server, remove all Perl scripts from the cgi-bin folder related to SSI Web 8.4.0, 8.4.2, and 8.4.4 and then restart Apache. This bug was introduced in 8.4.0.
  2. SSI Web was not handling certain network paths (UNC) properly. This caused SSI Web to display an error when reading local data. This was introduced in 8.4.0.
  3. Answer verification for Semantic Differential questions would fail if the left side of the scale was assigned a value greater than the right side of the scale.

The following bug fixes have been made in CBC:

  1. An error would be displayed if an attribute involved in an interaction was deleted and later you attempted to open the interactions area in the analysis settings.
Read More

08 July 2015

The following bug fixes have been made in SSI Web in general:

  1. Using the "Web server" option under the "Get Data" dialog could incorrectly merge respondent data for very large surveys. This was introduced in 8.4.2.
  2. In some situations questions that were pre-skipped inside of a loop would continue to be skipped on subsequent iterations of the loop. This was introduced in 8.0.0.

The following bug fixes have been made in ACBC:

  1. SSI Script in the Global Settings header or footer could prevent ACBC Test design from running.

The following bug fixes have been made in CBC:

  1. CBC HB and CBC Latent Class would only work if the respondent numbering was internal (sys_RespNum) or sequential. "Other" respondent numbering variables would not work. This was introduced in 8.4.0.

The following bug fixes have been made in MaxDiff:

  1. Direct binary logic in the settings for Anchored Maxdiff Analysis would only work if the numbers were decimal values. This was introduced in 8.4.0.
Read More

25 June 2015

The following improvements have been made in SSI Web:

  1. Downloading data is now faster and more efficient when downloading from SSI Web. This is most noticeable with large data sets.
  2. Find buttons have been added to help when looking for question anchors in loops and randomization.

The following bug fixes have been made in SSI Web in general:

  1. When starting SSI Web occasionally an error, "Can't locate perl518.dll" was displayed and SSI Web would fail to launch. This bug was introduced in 8.4.0.
  2. Pagebreaks would deselect if moved up or down in the Write Questionnaire dialog rather than continue to be selected. This was introduced in 8.4.0.
  3. In the Admin Module under the reports area decimal numbers were truncated into whole numbers. Now decimal numbers are not truncated.
  4. Free format numeric/open-ends would not auto restore answers when backing up if the type attribute of the input tag was different from 'text'.

The following bug fixes have been made in ACBC:

  1. ACBC merged rows would still display rows that should have been merged. This was introduced in 8.4.0.
  2. The ACBC alternative specific attributes dialog would sometimes show the wrong attribute when the summed price was not the last attribute.

The following bug fixes have been made in CBC:

  1. CBC HB and CBC Latent Class would only work if the respondent numbering was internal (sys_RespNum) or sequential. "Other" respondent numbering variables would not work. This was introduced in 8.4.0.

The following bug fixes have been made in MAXDIFF:

  1. Direct binary logic in the settings for Anchored Maxdiff Analysis would only work if the numbers were decimal values. This was introduced in 8.4.0.
Read More

03 June 2015

The following improvements have been made in SSI Web:

  1. SSI Web is now faster and more efficient. Data accumulate, export, and analysis perform faster, noticeably so with large data sets.
  2. Auto upload has been improved to be compatible with more FTP server software.
  3. Test options are now available on every page of the survey when you are in test mode. This allows you to change settings on any page. For example you can skip to any question in the survey from any page that you are on when you are in test mode.

The following bug fixes have been made in SSI Web in general:

  1. The license agreements for subscription and academic subscriptions have been updated. You may view license agreements at http://www.sawtoothsoftware.com/agreements.
  2. In some situations, backing up in a survey did not restore the prior answers. This could occur if the server was running Perl 5.18 or greater and other combinations of survey settings were in use.
  3. Updates have been made to reduce the number of errors being reported in the Admin Module error log. If the studyname being passed in the URL is wrong the respondent will see an error but it will not be recorded in the error log. Often these types of errors were generated by automated web crawlers. Additional updates have been made to reject automated web crawlers (spiders, robots) from entering the survey.
  4. In prior versions there was an additional system stylesheet for mobile. In this version the base mobile style settings have been included with the base style sheet. This allows users to more easily override the mobile CSS if they desire.
  5. Perl processes on the server will now be automatically terminated after 3 minutes. This will help to control run away Perl processes that could cause server performance issues.
  6. Certain fields starting wtih "sys_" were ignored for respondent duplication checks during data accumulation. They are now included in duplication checks.
  7. In the Web Server Management window, clicking "Get Links" without first uploading files would sometimes crash SSI Web.
  8. If the FTP settings file was corrupt, then SSI Web would not allow you to enter the Web Server Management dialog.
  9. SSI Web would crash if accessing the Web Server Management window without an internet connection.
  10. When an error occured during data accumulation, the line number reported with the error was off by one.
  11. Carriage returns (new lines) are now removed from open end data.

The following bug fixes have been made in ACBC:

  1. When using the ACBC Interaction Search Tool with a utility file for main effects, it was possible for an error to be displayed and the search would terminate.
Read More

06 January 2015

The following bug fixes have been made in SSI Web in general:

  1. Backing up over a question that had been answered and then skipped over would cause error 308.
  2. The JavaScript function SSI_GetValue would return numbers as text strings for both radio and checkbox questions. As a result numeric comparisons would not always evaluate correctly. SSI_GetValue now correctly returns numeric values for both radio and checkbox questions.
  3. When the End Unverified token was missing from constructed list logic, the verification error would say unexpected error instead of End Unverified missing. This was introduced in 8.0.0.

The following bug fixes have been made in CBC:

  1. After estimating utilities with HB with attribute coding set to Linear (HB Settings, Attribute Coding) an incorrect .val file would be written to be used in SMRT.
  2. Counts banners were being calculated incorrectly when there were overlapping category labels.
  3. The SSI Script function CBCDesignLevelValue would return "256" for the level if an attribute was missing (e.g. because of an alternative specific design). It now returns "0" if the attribute is missing.

The following bug fixes have been made in ACBC:

  1. During HB Estimation, if the attributes list was a constructed list and a particular level of the constructed list (an attribute) was never shown to any respondents AND that attribute was excluded from estimation, other included attribute utilities would be set to zero.
  2. During HB Estimation, if some levels had no level text, "The parameter is incorrect" would be shown. We now put a ":" placeholder for blank levels.
  3. ACBC None Threshold calculations would cause an error if any alternative-specific conditional attributes have attribute coding set to Linear.
  4. Price Adjustments could not be used if skipping the BYO. They are now allowed.
  5. ACBC Must Haves would sometimes return rules that didn't make sense for sequential attributes with no preference order if those attributes were not shown in the BYO.
  6. The documentation for the SSI Script functions ACBCWinnerLabel and ACBCWinnerValue was incorrect. The documentation said that the BYO concept would be returned if there was no tournament. This was incorrect. Nothing is returned if the tournament is skipped.
Read More

01 October 2014

The following bug fixes have been made in SSI Web in general:

  1. The SSI Script [%SetValue(varname, QuestionName())%] would fail. QuestionName() was being called twice and the second time it would return empty.
  2. You can now download either "All Data" or "Qualified / Complete" data only via the Admin Module.
  3. SSI Web auto upload now supports TLS connections and FTP Over SSL Modes explicit and implicit.
  4. SSI Script functions Value( ), GetValue( ), and SetValue( ) now verify properly with Free Format Hidden, Free Format Radio, and Free Format Select (Combo Box) text variables. Free Format Hidden, Free Format Radio, and Free Format Select (Combo Box) text variables now properly give an error during study verification when using functions AIELoopList( ), ANELoopList( ), AIGLoopList( ), and AILLoopList( ). Free Format Radio variables now verify properly with the function RadioSelect( ). The function parameters are now checked for matching data types. Text/HTML questions, Grid questions, and Free Format questions now work properly when used in the function Shown( ).
  5. The Shown() function would return true if any part of a question was shown even if a specific variable was passed in. For example with a Grid question G1 the SSI Script Shown(G1_r3) would return true if any row in G1 was shown, even if row 3 was not show. This has been fixed. See updated documentation on Shown().
  6. If a study had more than 8192 data fields and randomized pages, then an accumulation of data into SSI Web would fail.
  7. Limbo cleaning of data is now only triggered if changes have been made to answers.
  8. Export would display an error when a randomized block set was using a constructed list and none of the blocks were assigned to specific pages.
  9. Time stamps for errors in the error log were not in the same time zone as the data record time stamps. Now errors are in GMT time which is the same as the data records.

The following bug fixes have been made in CBC in general:

  1. Interactions defined in HB should have been ignored when running the Interaction Search Tool.

The following bug fixes have been made in ACBC in general:

  1. ACBC Interaction Search Tool with Summed Price and Individual Level Utilities is now available.
  2. If a nested conditional attribute contains multiple primary attributes, then in some design concepts that attribute could contain missing levels when there should be a value.
  3. SSI Script entered in ACBC Merged Row Labels would not execute.
  4. Introduced in 8.0.0 ACBC BYO questions on the first page of studies did not check for errors during study verify.
  5. An ACBC study that contained only conditional attributes (because all the primary attributes had been removed from the constructed list) would show a single screener question with no attributes or levels.
  6. The functions ACBCIsMustHave() and ACBCIsUnacceptable() did not execute correctly and would show an error when placed in a survey.
  7. The ACBC counts report now works with sequential numbering as the respondent numbering variable.
  8. ACBC exercises using an attribute that is both sequential and has a preference order, could introduce decimal numbers into the level count saved data if a respondent chooses an unacceptable rule that would violate the BYO chosen level for that attribute. ACBC Counts Report would not run when this was the case.
  9. Rules applied from respondents in a Must Have question in an ACBC exercise containing only one Must Have question would still be seen in future unacceptable questions as duplicate possible rules that should not be seen. The extra rule could be chosen but not change the design on the fly as the design was already changed from the Must Have.
  10. When using constraints in ACBC with some attributes excluded, the constraint would be applied to the incorrect attribute.
  11. The function ACBCNumChoiceTasks() would not work if you only have 3 concepts (including BYO) going into the tournament. Now if the number of concepts going into the tournament matches the number of concepts shown per task then 1 task is shown.
Read More

22 May 2014

The following bug fixes have been made in SSI Web in general:

  1. In version 8.3.4 a bug was introduced having to do with detecting the correct I.P. address of the respondent. The bug would trigger the error message "Error #274 - Database error. Cannot insert history row". This error would prevent respondents from taking the survey. Only some server configurations (those that included proxy servers and load balancers) were affected by this bug.
  2. When opening up two instances of SSI Web a error message was displayed and local testing (local web) would turn off.

The following bug fixes have been made in CBC and ACBC:

  1. With certain studies the Interaction Search Tool would display an out of memory error.
Read More

15 May 2014

The following bug fixes have been made in SSI Web in general:

  1. An error would appear when trying to delete custom survey styles.
  2. Error 302 gets generated when a respondent uses a mid-survey URL with a different browser or IP address than originally used. This error still gets generated, but it is no longer reported in the Admin Module error log.
  3. Error 303 has been removed. This was generated when mid-survey links expired after 2 hours.
  4. In some situations grid questions inside of loops would not show up as options in the Admin Module reports area.
  5. Pre-skips to "End of Block" inside of randomized blocks did not work correctly.
  6. CSV exported timestamps now include seconds.
  7. Questions with only 1 letter names would not verify in SSI Web if they were included in loops and referenced in constructed list logic outside of the loop.
  8. Backing up to a page with a pre-skip would in some cases not skip the original questions that were skipped.
  9. If a study has more than 8K data fields when exporting to csv, SPSS, Excel, or tab/lay, then free format variables would cause the export to fail completely.

The following bug fixes have been made in MaxDiff:

  1. Unverified Perl was not executing if entered into the MaxDiff "Response Options Text" area inside Anchored Scaling.
  2. Version 8.3.0 introduced a bug where the prior covariance matrix used in MaxDiff/HB was not correct. This does not appear to have a significant effect on estimated utilities.

The following bug fixes have been made in CBC:

  1. Running HB would run out of memory when generating the spreadsheet report if enough parameters and respondents were analyzed.

The following bug fixes have been made in ACBC:

  1. In most cases price adjustments were not being displayed correctly after the BYO question.
  2. If multiple price adjustments were applicable to the same attribute and level then only one of them was applied instead of aggregating them. This occured after the BYO question.
  3. When ACBC Interaction Search Tool was run on a study with Summed Price, and an .hbu file was specified, the results were not correct. This scenario is prevented in this version of the software and will be fixed in the next release of SSI Web.
  4. The SSI Script function ACBCNumChoiceTasks() would return 1 even if there were no ACBC Choice Tasks shown.
  5. When ACBC has many prohibitions and a valid BYO cannot be found, ACBC Test Design would throw an unhandled exception.
  6. Previous to this version dropped levels could not be estimated as 'Inferior' correctly when the BYO section was not included in estimation.
  7. Previous to this version when running Otter's Method constraints would cause the estimation to fail to converge or result in very flat estimates.
  8. Version 8.3.0 introduced a bug where the gammas output file for Otter's Method included the same value for all three columns.
Read More

25 March 2014

The following bug fixes have been made in SSI Web in general:

  1. An error was generated when constructed lists containing "Other Specify" data was used in either Select combo boxes or Grid combo boxes.
  2. The JavaScript function SSI_GetValue() would remove leading 0's from open ended responses that contained numeric input.
  3. Large decimal point numbers entered as numeric input would produce an error with Windows SQL Server.
  4. Numeric data fields would sometimes come out as scientific notation if they were small enough. This caused issues when exporting to SPSS and data collection with SQL Server.
  5. If open ended questions contained only numbers, then they were being truncated when brought into SSI Web to 5 decimal places.
  6. If open ended questions contained only numbers with leading zeros, then clicking the previous button would not restore the answer. In addition the leading zeros were not imported into SSI Web.
  7. Clicking cancel in the data generator while it would build the local test folder causes the data generator to display various error messages and then not work.
  8. The "Delete Survey from Server" button did not remove the "url_restore" database table.
  9. If Page Time was turned off then backing up via the previous button to a text/html filler question did not work correctly. The respondent would go to the question prior to the text/html filler question.

The following bug fixes have been made in CVA:

  1. The R-squared value in CVA HB has been updated to be between 0 and 1 just like other analysis areas.

The following bug fixes have been made in MaxDiff:

  1. Sequential Respondent Numbering did not work for MaxDiff HB, CBC Counts, or ACA acd export.
  2. The Respondent Numbering setting was not always used in analysis.
  3. The MaxDiff SSI Script functions MaxDiffDesignValue() and MaxDiffDesignLabel() produced a script error when called. This was introduced in 8.3.0.

The following bug fixes have been made in CBC:

  1. Interaction Search Tool would not function.
  2. Most cases of conditional display and alternative specific attributes together did not work properly.
  3. CBC Counts report could run out of memory in some cases when using 3-Way Interactions.
  4. Data Management would not display numeric values with decimals.
  5. CBC Constant Sum/HB would not run if respondents did not answer all tasks. Now any answered tasks are considered.
  6. CBC/HB will give more user friendly error messages when it fails.

The following bug fixes have been made in ACBC:

  1. A constructed attribute list that contained an exclusive item was not being handled in analysis for alternative specific relationships and merged rows.
  2. Restore defaults in ACBC Analysis wasn't resetting all the settings to default.
  3. ACBC Constraints did not allow the same value for a Preference Group in a single row for Relative Preference Groups.
  4. Conditional Display used together with Merged Rows would not properly remove blank items when displayed in the web browser.
  5. The BYO did not display correctly (in some instances) nested Alternative-Specific relationships.
  6. ACBC BYO can now be skipped even with a summed price attribute.
  7. The none threshold point of purchase intent was not allowing decimals.
  8. If am empty data file existed when studies were converted, then Data Management would not work.
  9. Conditional Attributes could still in rare cases exist in designs when they should not.
  10. ACBC HB will not halt if a respondent does not have task data and will just ignore the repondent.
  11. ACBC Summed Price was broken if Avoid Dominated Concepts was unchecked.
  12. Data Management was truncating the display of text values to 255 characters.
  13. ACBC Unacceptables/Must haves now trim other rules based on conditional attributes when all their primary attributes and levels have been marked as unacceptable.
  14. ACBC Test Design would crash if too many prohibitions existed and no designs could be created.
  15. If all primary attributes were dropped from ACBC using constructed lists, then the data collection would give an erroneous error.
  16. ACBC alternative relationships and merged rows did not work when using exclusive ("none of the above") items with constructed lists.
  17. ACBC Must-haves/Unacceptables will only consider the BYO task for creating rules if respondents saw the BYO.
  18. If the ACBC price attribute was dropped using constructed lists and the BYO was not shown then ACBC designs were incorrect.
  19. In ACBC/HB, if all BYO attributes were unchecked but the top BYO node was checked, then HB would not run.
  20. Accumulating only merged variables normally auto combines the new data with existing records. If any variables were marked ignored, then it prompted the user about what to do. Now the default behavior is applied.
  21. Right clicking the View/Data Sheet when no data exists was giving a strange error message.
  22. If ACBC only used a small number of attributes and levels, it was possible the BYO concept would be created as a screener concept.
Read More

19 February 2014

The following features have been added to ACBC:

  1. Alternative-Specific Designs - Users can now specify that certain attributes (with all their levels) are only applicable to certain level (e.g. brands, styles, etc.) of a primary attribute. For example, if comparing desktops to laptop computers, the laptop has a battery whereas desktops do not. There may be a conditional attribute upon the presence of battery, such as three levels of battery life. With alternative-specific designs, the attribute Battery Life (with its three levels) would only occur with laptop computers.
  2. Merged Rows - Especially with Alternative-Specific designs, it is now possible to join or merge the level text from two or more attributes.
  3. Ability to Skip the Screener Section - It is now possible to skip the Screener section, meaning that the respondent can jump straight from the BYO question to the Choice Tournament, or right into the Choice Tournament (skipping both the BYO and Screener sections).
  4. Improved Design Efficiency - ACBC designs are significantly more efficient than before. The designer goes much further to increase the orthogonality of the experiments at the individual-level. Our tests suggest that the D-efficiency (at the respondent level) has improved by about 5%-40%, depending on the attribute list and questionnaire setup.
  5. Test Design Routine - ACBC Test Design generates a few or hundreds of "robotic" respondents (where the respondents answer randomly) and provides key statistics regarding the quality of the designs: level frequency, balance, d-efficiency, and standard errors (precision) of the utility parameters.
  6. Avoiding Dominated Concepts - A dominated concept is a product alternative that is logically inferior to another concept (based on the Preference Order settings the researcher specifies). You now have the ability to specify whether to avoid dominated concepts.
  7. Ability to Customize the "At Most" and "At Least" Text by Attribute - Must Have questions now allow you to specify "At Most" and "At Least" text per attribute. In prior versions the same text was used for each attribute.
  8. Interaction Search Tool - A new aggregate logit utility estimation procedure is available to help researchers investigate which interaction effects are potentially most significant.
  9. BYO Concept Included in Choice Tournament The option is now available for the BYO to be included in the Choice Tournament.
  10. Exclude Certain Attributes from "Must Have/Unacceptable" Questions - The option is now available to exclude certain attributes from the Must Have/Unacceptable questions.
  11. Covariates in ACBC/HB Utility Estimation - The same capabilities that CBC/HB has regarding the use of covariates now exist in the HB routine for ACBC.
  12. More Flexible Utility Constraints During Utility Estimation - The option is now available to specify that certain levels must be preferred to other levels within an attribute. Multiple independent chains of ordinal constraints within an attribute can now be specified.
  13. Record and Export Tournament Concept Order - The way the Choice Tournament data is saved has been changed. It is now possible to detect whether or not respondents are always answering the first concept etc.
  14. Conditional Display Table Allows for "Missing" Levels - If you have an Alternative-Specific design in ACBC, and you are using Conditional Display, then it is now possible for you to specify what is displayed when a missing level occurs. The 0 level in the table represents the "missing" (not applicable) level.
  15. Price Levels ToolTip - A tooltip has been added to help know the default price levels. These values were previously calculated once and then lost when the settings were changed.

The following features have been added to CBC:

  1. Interaction Search Tool - A new aggregate logit utility estimation procedure is available to help researchers investigate which interaction effects are potentially most significant.
  2. Option to Constrain Groups to a Common Scale - A new option for latent class estimation has been included. Now in each iteration of the latent class procedure, the scale factor for the part-worth utilities across groups is constrained to be equal to the average scale factor for the groups. This constrains latent class from returning solutions where two groups differ only in terms of scale factor (response error).
  3. Not Applicable (Missing) Levels in Conditional Display - If you have an Alternative-Specific or Partial Profile design then it is possible for you to need to specify in a conditional display table what is displayed when a missing level occurs. The 0 level in the table represents the not applicable level.

The following features have been added to MaxDiff:

  1. Option to Constrain Groups to a Common Scale - A new option for latent class estimation has been included. Now in each iteration of the latent class procedure, the scale factor for the part-worth utilities across groups is constrained to be equal to the average scale factor for the groups. This constrains latent class from returning solutions where two groups differ only in terms of scale factor (response error).
  2. New SSI Script Function MaxDiffVersion() - This new function returns the design version number and question name. Useful when creating paper-and-pencil interviews.

The following features have been added to SSI Web in general:

  1. JavaScript Required by Default - By default all surveys now require the respondent to have JavaScript turned on in their browsers. If they do not have JavaScript they will see an error message prompting them to turn it on.
  2. Better Handling of Extremely Large Surveys - SSI Web previously could crash when surveys became extremely large, such as 60K data fields or larger. This occurs when preparing for web uploading, accumulating, exporting, or analyzing data. Large data sets can now be handled.
  3. Can Cancel Accumulating Data - You can now cancel the process of accumulating new data.
  4. Survey Links Can Now be Copied - Added a "Copy Links" button to Auto Upload to quickly copy survey links to the clipboard. There is now a "Get Links" link allowing you to see the links again without having to upload again.
  5. Merged Variables Upadate - Merged Variables changed to be Categorical, Continuous, and Text. Categorical variables can now be assigned category values and labels. These variables can now be used as HB Covariates.
  6. Set Starting Internal Respondent Number - You can now specify a starting point for the Internal Respondent Numbers. See Server Settings | Advanced | Respondent Number Start.

The following bug fixes have been made in SSI Web in general:

  1. SSI Script function RadioSelect() now accepts numbers or text as arguments.
  2. Questions named the same name as certain JavaScript keywords (such as "constructor") would cause JavaScript to fail.
  3. Prior to this release all surveys viewed in Microsoft Internet Explorer were displayed as if they were being viewed in Internet Explorer 7, even if Internet Explorer was version 8 or greater. This has been removed. This change also allows Windows Phone devices to display the mobile style sheet.
  4. The ToolTip function when viewed in the survey would use up an extra space at the end of the ToolTip link.
  5. In the Admin Module under Data Reports, the chart type was not being saved in a saved report. As a result the pie chart could not be used in a saved report.
  6. The Custom Handle Start Position in Semantic Differential quesitons did not work correctly.
  7. In certain situations with inner loops and constructed lists the error message "Circular reference discovered in list building causing an infinite loop" would be generated and the respondent would not be able to continue the survey.
  8. In the Admin Module the View Data option would not work if viewing more than about 2,300 records at a time.
  9. If a pass-in field was set as "Decimal Number", and the server was using MSSQL as the database, then restarts using a link with the pass-in field would not work.
  10. In some situations loops involving constructed lists would not work correctly and a error message would be displayed.
  11. The constructed list logic for AIL, AIE, AIG, and ANE, etc. has changed in the way that it considers empty values. Empty values are no longer considered in any of these commands. For example AIL(Q1, 5) will only return answered items less than 5. The documentation was incorrect on this matter and has been updated.
  12. Constructed lists that were created inside of loops, and used the RANDOMIZE() command, would have the same random order for each loop iteration.
  13. The respondent's I.P. address was being recorded in the history table even if "Save Respondent I.P. Address" was unchecked.
  14. Open-End type questions did not save new lines and didn't save trailing \ characters.
  15. The constructed lists used to drive randomized block order and loops was not saving properly in certain instances when used together.
  16. Not equal did not work when used in filters.
  17. Respondents who entered the survey using the bypass link were able to click the previous button which caused problems.
  18. Backing up in loops would not function properly when some pages in the loop are answered by the respondent and others are auto answered. It fails to back up at all or goes to the wrong page.
  19. If an empty randomized block set existed in a study with loops, then extra unnecessary data fields were created in the study. This caused SSI Web to slow down considerably.
  20. Randomized blocks were not being ordered properly when blocks were not set in page ascending order.
  21. When randomized blocks were non contiguous, meaning there are questions between the blocks that are not in any block, and pre-skips where used in certain cases, the block ordering would not save. This resulted in respondents not seeing the survey in the proper order.
  22. The Data Generator using browser mode on a remote server would not work on some studies.
  23. Value labels were not stripping out html for SPSS and layout file exports. This was introduced in 8.0.0.
  24. Auto Upload would error out when uploading over a previously existing version of SSI Web on a remote server.
  25. When SSI Web had accumulated data for numeric data fields that belonged to a question that no longer existed in the study, and that field was exported to SPSS, SPSS would error out with code 20.
  26. In CBC HB, ACBC HB or ACBC Monotone Regression, if linear, log-linear, or piecewise attribute coding was used, the importances in the report generated were incorrect. This has been in the software since version 7.

The following bug fixes have been made in CVA:

  1. When specifying anchor text for an exercise, if you opened the large text editor view for any of the labels, you would encounter an error. This was introduced in 8.2.0.

The following bug fixes have been made in MaxDiff:

  1. The fit statistic is now computed on a 0 to 1 scale for MaxDiff utilities.

The following bug fixes have been made in CBC:

  1. Running HB with a single respondent would show average utilities as zeros.
  2. Importing a design file in CBC was not working properly. The first design would be imported and then all remaining designs would receive the first design. This was introduced in 8.2.4.
  3. Decreasing the number of fixed choice tasks would not visually remove the excess task columns from the screen where the fixed choice tasks are defined.
  4. The automatic analysis utility exports added in 8.2.4 exported to the .xlsx file format. We learned that this format was not opening in spreadsheet tools other than Excel (OpenOffice, LibreOffice, Google Drive). Now, after a utility run, the .xls and .xlsx file formats will be exported automatically.
  5. Dual *.csv export was exporting the fixed tasks first in the design file, but in question order in the response file.
  6. In some situations Conditional Relationship text would not be displayed if the attribute was involved in both Null Level Prohibitions and Conditional Relationships and "Align Common Attributes on Same Row" was checked.
  7. Single *.csv exports had incorrect random task numbering.

The following bug fixes have been made in ACA:

  1. An ACA .acd export that included respondents with an incomplete pairs section would give the following error: "Failed. Input string was not in a correct format." This issue was introduced in version 8.0.0.

The following bug fixes have been made in ACBC:

  1. Single level attributes are no longer allowed in an exercise.
  2. SSI Script functions ACBCBYOLabel and ACBCWinnerLabel now show the formatted price for the summed price attribute.
  3. ACBC would encounter problems in analysis when the exercise only had 1 attribute. This is no longer allowed with predefined lists.
  4. A tooltip has been added to help users know the default price levels. These values were previously calculated once and then lost when the settings were changed.
  5. The Pricing Unacceptable label was not being exported for translation.
  6. An ACBC exercise with 2 attributes, and both were included in the BYO, would sometimes show the BYO as a screener concept.
Read More

05 June 2013

The following bug fixes have been made in SSI Web in general:

  1. When exporting any file in Windows XP, after the first action in the first selected job would finish, an "Out of memory" error would occur, no matter the size of the data being exported.
  2. The Shown() SSI Script function now supports text / filler questions.
  3. No warning was displayed while changing list items that were associated with Grid questions.
  4. Exporting a variable that was whole numbers for all respondents but was set to decimal would cause Error code 20 for SPSS exports.
  5. When copying Constant Sum Grid questions, the settings would be reset to the defaults.
  6. The SPSS settings for stripping HTML were not being saved when a survey was closed and reopened.

The following bug fixes have been made in MaxDiff:

  1. When importing Xliff files containing Indirect Anchored Scaling, the order of "All", "Some", and "None" was not preserved (First column would show "All", "None", "None") and the text that should have been for "Some" and "None" was reset to the defaults.

The following bug fixes have been made in CBC:

  1. Pressing delete in Conditional Relationships resulted in an error.
  2. Pressing the pencil icon in Conditional Relationships results in an error.
  3. The individual importance scores in the HB report were incorrectly displayed as equal to 100 divided by the number of attributes. This error was introduced in version 8.2.2 and has been fixed.
  4. Importing more prohibitions than there are rows would result in an "Index out of bounds" exception.

The following bug fixes have been made in ACBC:

  1. For ACBC Screener questions, setting the option of "Show Attribute Labels" to "No" was not removing attribute labels from tasks.
  2. The Data Generator would not work if the ACBC BYO Layout was set to Combo Box.
  3. ACBC Price Adjustments in the BYO question did not work correctly if combo boxes were used. The price at the end of the level text (i.e. + $100) would not change based on Price Adjustment rules. The amount in the total boxes was however correct. This bug was introduced in 8.0.0.
  4. Not enough data columns were preallocated in the database for ACBC in some rare conditions meaning the respondent could not continue answering.
Read More

23 April 2013

The following features have been added to SSI Web in general:

  1. A new SSI Script list function has been added named Mirror. Mirror is useful to map or mirror one list onto another when those lists have different parents.
  2. In the Admin Module there is now a way to download the raw database tables. Use the new "Advanced" link to access this advanced feature.

The following bug fixes have been made in SSI Web in general:

  1. If you entered the advanced area for a Ranking question, when you left the page, the normal Ranking question options window would close and nothing would be saved.
  2. Autoupload better handles server page redirects.
  3. When accumulating data and renumbering respondents, 3 additional system columns are created to help match the renumbered respondent with the original respondent. These additional fields were not exportable.
  4. Exporting Free Format hidden, radio, or combobox marked as decimal would truncates all decimal digits - This bug has been present since 8.0.2.
  5. The amount of data fields in some studies was incompatible with the maximum number of data columns per data table in MySQL. MS SQL had a similar issue. This error was seen upon setting up the study on the server or applying changes.
  6. If the data generator was launched and the study contains a Semantic Differential question with no assigned list for anchor list or left list or right list, then the data generator would not work.
  7. Data Generator breaks if the study contains any merged fields.
  8. Data Generator was generating commas for decimal numbers on a German OS.
  9. In the Data Generator if a defined value was specified for an ACBC Screener task, it would not work properly.
  10. In some languages, importing a csv file would result in the error message, "SQLite error AUTOINCREMENT is only allowed on an INTEGER PRIMARY KEY." This has been fixed.
  11. The Internal Respondent number can be set by adding respnum_start: (number here) to the end of the config.cgi file. This did not work with the MS SQL database.

The following bug fixes have been made in MaxDiff:

  1. The utility graph in v8.2 did not zero-center the utilities, which is done to assist in assessing convergence.
  2. If you converted a MaxDiff study from 8.1.4 and tried to run Latent Class analysis, you would see an unhandled exception.
  3. The utility graph in v8.2 would go directly to the utility report instead of remaining so users can evaluate the convergence of the solution.
  4. MaxDiff analysis would not allow a Prior Variance of less than 1.
  5. Accumulation from csv is now case-insensitive, but fields in view/edit data will match the study variables.
  6. RLH fit statistic in MaxDiff is now scaled between 0 and 1 instead of between 0 and 1000.
  7. Significant performance improvements were made from the 8.2.0 version of MaxDiff HB. This puts completion times on par with the 8.1.4 MaxDiff HB version if not slightly faster.

The following bug fixes have been made in CBC:

  1. Slight improvements were made to the Conditional Relationships area to allow for more attributes/levels to be involved in conditional relationships.
  2. Excluded Attributes were included when rescaling to zero-centered diffs.

The following bug fixes have been made in ACBC:

  1. Excluded Attributes were included when rescaling to zero-centered diffs.

The following bug fixes have been made in CVA:

  1. Excluded Attributes were included when rescaling to zero-centered diffs.

The following bug fixes have been made in ACA:

  1. Preview would not work correctly with ACA attributes and levels.
Read More

19 February 2013

The following features have been added to SSI Web in general:

  1. Data Generator
  2. Streamlined Sawtooth Software Hosting Setup
  3. Smarter CAPI Data Merging
  4. New Test Survey Option: Disable Terminate/Link
  5. Toggle Intelliprompt

The following features have been added to MaxDiff:

  1. Anchored Scaling
  2. Logit and Latent Class Integration
  3. Covariates within HB Analysis for MaxDiff
  4. (1 of 12) etc. Counting Functions for MaxDiff Questions

The following bug fixes have been made in SSI Web in general:

  1. The previous button in certain circumstances would not allow respondents to back up. This was introduced in 8.1.4.
  2. Data collection would trigger an error if a grid row or column list was empty but not both.
  3. There was a typo in the autocomplete functions "AreAllQuotaClosed" and "AreAnyQuotaClosed". They have been corrected to be "AreAllQuotasClosed" and "AreAnyQuotasClosed".
  4. If passwords and cookies were used together then the cookies would not work. Some people might do this if they had multiple respondents per password. This bug has existed since v8.0.0.
  5. Open end responses inside of double quotes could be removed if the respondent backs up in the survey.
  6. Constant Sum questions and numeric Grid questions, where the total is displayed, might not always sum up correctly if using decimal point numbers.
  7. SSI Web was incorrectly auto-downloading data from the Admin Module in a non-UTF8 compatible manner. This caused some international characters to be imported incorrectly.
  8. Some export job settings were not saving correctly.
  9. SSI Web would not continue accumulating data when some respondents were marked to be overwritten.
  10. Loading export fields in an export job with lots of fields took several minutes. Any changes also took several minutes. This has been reduced to just a few seconds.
  11. Pressing enter before a block of SSI Script at the very beginning of the text editor would cause a crash. Navigation keys would sometimes produce unexpected behavior, such as disappearing text.

The following bug fixes have been made in CBC:

  1. CBC Best Worst value labels were incorrect.
  2. Running CBC/HB Analysis with a custom prior covariance matrix would result in an error.
  3. If a CBC exercise contained a dual response none, SSI Web would not be able to estimate utilities in HB if include none was unchecked.
  4. Some segment variable logic would cause SSI Web to crash.
  5. If CBC Export file type was set to dual response CSV and the Traditional None Option was set, then imported into the CBC/HB standalone, the none option was not being exported correctly.
  6. Some spreadsheets that are written during estimation were protected by default in Excel. All sheets are unprotected now.
  7. Unverified Perl would not be properly executed if entered into CBC Conditional Relationship text.
  8. When estimating utilities, where one or more of the attributes were excluded, using CBC/HB within SSI Web, the labels file (*.xml or *.labels) that is created with the *.csv file of utilities was not compatible with the Sawtooth Software Online Simulator. An error was shown when trying to upload the labels file to the simulator and the upload would not succeed.

The following bug fixes have been made in MaxDiff:

  1. MaxDiff *.cho export was ignoring the selected respondent numbering field.
  2. MaxDiff *.cho export with constructed lists failed. This bug has existed since 8.0.0.

The following bug fixes have been made in SSI Web CAPI:

  1. Get Data would have troubles downloading a CSV that was compatible with some international characters.
  2. Get Data and Load CAPI File would not work if the local server was not running.

To update SSI Web CAPI go to: http://sawtoothsoftware.com/resources/software-downloads/ssi-web-capi

Read More

03 December 2012

The following bug fixes have been made in SSI Web:

  1. Auto upload would time out when a study took a long time to build the web upload folder. This has been fixed.
  2. Auto upload would fail on linux if the ftp user name contained any of the following legal characters: `~!.()[]?
    This has been fixed.
  3. A bug was found in Random Blocks. If the order of the blocks was determined by a list, and the list was empty, a server error would be generated. This has been fixed.

The following bug fixes have been made in CBC:

  1. When running CBC/HB with covariates, demographics were not being coded into the proper covariate vector.  This has been fixed.
  2. Fixed a bug in the SSI Script function CBCAttributeLabel(). This function would not execute nested functions. This has been fixed.

The following bug fixes have been made in MaxDiff:

  1. The last respondent record was not being written to the csv output file for MaxDiff HB Scores. This has been fixed.
Read More

22 October 2012

The following bug fixes have been made in SSI Web:

  1. If a quota question was on the same page as other questions, and the quota question was at the top of the page, then data on that page would be marked as limbo data and removed from the actual data file. This bug was introduced in 8.1.0 and has been fixed.
  2. Exporting to tab/lay files with blank time stamps for start and end times would not write just empty spaces but nothing at all. This would cause the spacing counts to be offset. This has been fixed.
  3. Accumulating would mark some empty fields as differences as well as identical numbers. Data like 3 and 3. were marked as differences when they shouldn't have been. Two blank fields are no longer marked as differences. This has been fixed.
  4. Studies containing a free format question in a loop would fail during export when a recode map was applied to the export job and the free format question was selected for export. This has been fixed.
  5. Auto uploading with bigger studies would not wait for the web upload folder to finish building. This was introduced in 8.1.0 and has been fixed.
  6. Some rare circumstances of auto upload could cause it to never terminate. This was introduced in 8.1.0 and has been fixed.

The following bug fixes have been made in ACBC:

  1. A bug was fixed in ACBC analysis (utility estimation and .cho export). The bug only affected studies that use a constructed list for the attribute list and have one or more attributes with the "Include in BYO" option set to "No". The problem affects proper utility calculations and also design information in an exported .cho file. BYO data recorded in the data file are not being read correctly during ACBC analysis. As a result, an incorrect design for the BYO question is used for utility estimation or exported to the .cho file. This bug was introduced in version 8.0.0 and has been fixed.

The following bug fixes have been made in CBC:

  1. In a CBC Exercise, changing the number of concepts was not properly reflected in the Task Labels tab. This has been fixed.
  2. Dual response coding is wrong when using worst tasks only.  This has been fixed.
  3. Dual response information is thrown away when using best or worst tasks only.  This has been fixed.
  4. The CBC Counts Report was being saved with the incorrect extension. This has been fixed.
  5. In CBC Latent Class the sequential respondent number was always written to the *.pxx files instead of the selected respondent number. This has been fixed.

The following bug fixes have been made in MaxDiff:

  1. The Max Diff Counts Report would fail with error message "Error saving Excel file" when exporting to Excel. This has been fixed.
Read More

24 September 2012

The following features have been added to SSI Web in general:

  1. We've improved the rendering of SSI Web surveys on mobile devices. If we detect that the screen resolution is less than 600 pixels wide, we include a mobile CSS style sheet that makes the text and buttons more readable and clickable on mobile devices. Check out some mobile screenshots showing "before" and "after" illustrations of what we've accomplished.
  2. The Online Administrative Module has a new mobile interface that allows mobile users to monitor the progress of their surveys and change quotas more easily on a small screen. You can still access the standard Online Administrative Module using your mobile if you want to access the full capabilities of the standard management module.
  3. The Text Editor has been improved. HTML attribute and value prompting, auto-indentation on enter, Javascript and CSS prompting, and word wrap toggle.
  4. Auto Upload now gives you the study links after uploading is complete.
  5. We have updated the SSI Web help on our web site. See http://www.sawtoothsoftware.com/help

The following features have been added to CBC:

  1. Best-Worst choice format option. Respondents can select best AND worst concepts within each task. It isn't clear from recent research that collecting both best and worst choices is superior to (time-equalized) best-only choices, but there has been an increase in interest for this and some research papers have shown respectable results.
  2. Counts and aggregate Logit available within the SSI Web menus. No need to export the data to the separate SMRT package to do the Counts and Aggregate Logit analyses.
  3. CBC/HB and Latent Class available within SSI Web menus. You may run estimation directly within SSI Web's Analysis menu.
  4. Concept labels. Now you can place labels over the concept columns, such as "Option A," "Concept 1", etc.
  5. Countdown text. The text "(1 of 10)" "(2 of 10)" etc. is automatically added to the header of each choice task, informing respondents how many tasks they have completed and how many more are remaining. You can delete this notification or modify it if you wish.

The following features have been added to MaxDiff:

  1. Improved Experimental Designs. We've improved the algorithm for selecting designs, leading to near-perfect within-version level-balanced designs. This leads to slightly improved HB scores estimation. You can read a white paper describing the designer improvements at: http://www.sawtoothsoftware.com/download/maxdiff_designer_v8_0_4.pdf

The following features have been added to ACBC:

  1. "GetPrices" Routine within SSI Web interface. We've integrated this tool for tabulating the frequency of summed prices shown across all respondents within SSI Web's Analysis menu. It previously was available as a separate .exe. This tool helps guide the choice of cutoff points in piecewise (non-linear) price function specification.
  2. Improvement in HB estimation when attributes are dropped via constructed lists. The old approach could lead to unusual results when the majority of respondents dropped an attribute as completely unimportant and a minority of respondents thought the dropped attribute was critical.

The following bug fixes have been made in SSI Web:

  1. Some improvements have been made to recording page time. Respondent Time per Page is now measured in seconds instead of milliseconds.
  2. Removing a study with the "Delete Survey from Server" button could remove files and folders that were not uploaded by SSI Web. This has been fixed.
  3. When recoding a loop list, the exported variable labels did not reflect the recoding of the loop list. This has been fixed. For instance, if the list used by a loop is recoded during export to make it 0-based, and the question "FavoriteColor" is in that loop, the label "FavoriteColor.1" is now recoded to "FavoriteColor.0".
  4. When adding prohibitions, an "Index out of range" error would sometimes occur and prevent the addition of further prohibitions.
  5. Sometimes a database error #203 would be generated by respondents backing up in the survey. This has been fixed. This was a problem with MSSQL database servers.
  6. In Test Mode in the Admin Module, if a particular respondent number was specified, and there was already data for that respondent number, constructed list data would not be cleared out from the prior test record.
  7. Empty constructed lists could produce a server error in some cases. This has been fixed.
  8. The .labels file export would not escape XML entity characters properly. This has been fixed.
  9. When exporting an empty start or end time to spss or tab/lay, the export would fail. This has been fixed.
  10. Sliders in Semantic Diff could be inaccurate if "Use Scale Anchor Text List Values as Saved Values" was selected and the number of items in the list was 21, 31, 41, etc.
  11. Sliders did not work on mobile devices. This has been fixed.
  12. Open end data with a decimal point number in it would not be displayed correctly in the Admin Module View Data area.
  13. If Free Format hidden variables were set as "Numeric", and a decimal point number was assigned to it, only the whole number portion of that number was saved. This has been fixed.
  14. Merged data could not be imported if there were newlines in an individual data field.
  15. Importing more than one xliff file into the same study would not behave properly. After the first import, any modifications would not be brought into the study and the original text would remain. This was introduced in 8.0.0.
  16. In the Data Management area, when renaming a column the data in that column would be deleted. This has been fixed.
  17. If a Tool Tip was on the far right of the screen the pop-up text would flow off the screen. This has been fixed.
  18. When exporting to SPSS, variable names would fail (SPSS Error Code 13). This was introduced in 8.0.0 and has been fixed.

The following bug fixes have been made in ACBC:

  1. ACBC creates designs by changing up to half of the attributes specified in the BYO question. If the number of attributes was even then one more than half were changed. This has been fixed.
  2. Having only SSI Script in ACBC pricing fields would cause the ACBC Analysis Settings dialog to enter into a loop with the error "Please enter at least two price values." This has been fixed.

The following bug fixes have been made in CBC:

  1. With CBC Constant Sum, the last item in the task is not recognized as a variable in SSI Script. For instance, if there were 8 concepts in the first CBC Constant Sum task, using CBC_Random1_8 in skip logic would not pass the "Study Verify" checks.
  2. When moving attributes around in a CBC exercise, there was a case where moving them in certain ways would result in an "Index out of range" error.
Read More

27 April 2012

The following features have been added to SSI Web:

  1. A new text editor has been added to SSI Web. For more information see the blog post.
  2. The Auto Upload dialog (Field | Web Server Management) has been simplified. The Database tab and the FTP tab have been combined into one tab. Settings imported from 8.0.0 will need to be checked over.

The following feature has been added to ACBC:

  1. When calculating ACBC utilities, analysis constraints left blank are not constrained. For example, entering <2, 1, blank> as the relative preferences of a three-level attribute on the constraints dialog will cause only the 1st and 2nd levels to be constrained. Although this feature has been in the software since ACBC shipped, it has not been documented.

The following bug fixes have been made in SSI Web:

  1. There was a bug involving respondent passwords not being fully loaded into the database. If the STUDYNAME_pwds.csv was greater than 488 KB then it was possible that not all of the respondent passwords would be loaded. This bug could be triggered by manually loading the study via the Admin Module or by clicking Reset Survey in the Admin Module. This was introduced in version 8.0.0 and has been fixed.
  2. A bug in filter logic could cause incorrect logic being applied to null data values in the database when using the logic BLANK and !=. Changes were made to SSI Web and the Admin Module. Null is now treated the same as empty.
  3. The "Add Additional Filters" link in the Data Reports area of the Admin Module would not work correctly. An error message was displayed. This has been fixed.
  4. The built-in SSI_GetValue() JavaScript function would not work correctly when used on a text data field. This would cause a JavaScript error and JavaScript would fail where it was used. This bug was introduced in version 8.0.0 and has been fixed.
  5. SPSS variable labels were not being created by Semantic Diff questions.
  6. SPSS start and end times were converted to incorrect dates. This bug was introduced in version 8.0.0 and has been fixed.
  7. SPSS value label values for grid radio/combo values were zero-based.
  8. SPSS value label labels for grid radio/combo values were incorrect labels. This bug was introduced in version 8.0.0 and has been fixed.
  9. When using recode during export, Grid question radio values were not being recoded. This has been fixed.
  10. Version 8.0.0 would import the version 7 FTP settings and delete them from the hard drive so they no longer appeared in version 7. Version 8.0.2 imports them if needed but leaves the version 7 settings intact.
  11. When creating/editing export jobs in the export fields area, the categorizing of conjoint/maxdiff data fields was still happening with the original conjoint/maxdiff exercise when the exercise no longer referenced them. These fields are now marked as unknown.
  12. Multi-line other specify input boxes would not allow spaces to be typed. This was introduced in version 8.0.0 and has been fixed.
  13. Editing data in the View/Edit area of the Admin Module could fail and display an error message. This has been fixed.
  14. There was an error involving loops and building lists outside of the loop with data collected inside of the loop. For example AIC(Q1.1) would not behave correctly if Q1 was inside a prior loop and had a constructed list associated with it. This has been fixed.
  15. Semantic Differential questions were not allowed as parameters in various SSI Script functions. This has been fixed.
  16. Free Format radio, select, and hidden variables can now be identified as either Text or Numeric. Previously, radio and select variables were assumed to be numeric values and hidden variables were assumed to be text values.
  17. When accumulating data into SSI Web with merged variables, the combine feature of merging data did not function properly
  18. Free format hidden variables were not restored properly in some cases when the free format hidden value contained bracket characters.
  19. A bug was fixed that had to do with cross tabs and multi-selects in the Admin Module. Crossing a multi-select question with another question could result in an incorrect N at the bottom of the column variable. This could happen if one or more respondents saw the column variable but not the multi select.
  20. In some cases the export field order would not save properly in export jobs. This has been fixed.
  21. Backing up to a question that had been auto answered would back up to the question prior to the auto answered question and not restore any answers. This was introduced in version 8.0.0 and has been fixed.
  22. Saving to Excel or a *.csv on the analysis dialog would save locale-specific decimal numbers in the files. Now it will always use periods for the decimal point.
  23. Microsoft SQL Server would not handle single quotes entered into open end fields properly. Respondents would see a server error. This has been fixed.
  24. In the Data Management dialog's View/Edit tab, only two decimal places of precision were shown (with rounding). If a change was made to any data in the grid and then saved, the saved data were the numbers that were shown in the grid, a loss of precision for data with more than two decimal places. The new behavior is to show and save up to 16 decimal places of precision.

The following bug fixes have been made in ACBC:

  1. When calculating ACBC utilities, analysis constraints with the value of 0 were being treated as not constrained. For example, entering <0, 1, 2> as the relative preferences of a three-level attribute on the constraints dialog would only cause the 2nd and 3rd levels to be constrained. The same would hold true for constraints customized based on other survey data (i.e. using a question name for the constraint). This bug has been fixed. A notice containing more information sent to ACBC users is available by clicking here.
  2. The ACBC Tournament section's option to Show Alternating Colors was not working correctly. This was introduced in version 8.0.0 and has been fixed.
  3. If ACBC analysis constraints are used, leaving one or more cells blank when specifying the constraints will cause an error while computing utilities. The error reports that a column does not exist. This bug was introduced in version 8.0.0. It has been fixed.

The following bug fixes have been made in CBC and CVA:

  1. CBC and CVA Conditional Text items were limited to 256 characters. This was introduced in version 8.0.0 and has been fixed.

The following bug fixes have been made in CBC:

  1. The .cho export did not handle more than 4 conditional price attributes. Now it will, while displaying a warning that SMRT cannot read the .att file
Read More

13 February 2012

The following features have been added to SSI Web:

  1. Side-by-Side Install - Version 8 may be installed on the same machine as earlier versions.
  2. Database - Data is now stored in a database file on the server. MySQL and Microsoft SQL databases are supported.
  3. Crosstabs and Charts in the Admin Module - Reports involving multiple questions may be created and saved. The reports may employ both cross-tabulation and charting.
  4. Looping - This feature adds the ability to ask a question (or series of questions) multiple times, for each element of a list. Loops within loops are also supported.
  5. Data Management and Exporting Features
    1. View/edit Data allows you easily to see and edit the data similar to how you might view or edit data in a standard spreadsheet table.
    2. Multiple export "jobs" can be saved, so you can manage multiple export formats and lists of questions to be exported.
    3. Advanced respondent filtering logic can be used when exporting.
    4. During data export, you can recode (re-assign) the numeric values that are associated with list items. This makes it easier to create data exports that satisfy the particular demands of your client.
    5. Previously, open-end answers could be exported as grouped by question. Now, you can also export open ends grouped by respondent.
  6. Semantic Differential Question - This is a new question type where respondents use a sliding scale (from left to right) to make a choice between two words, such as "friendly" vs. "unfriendly". Both sliders and radio button scale options are supported.
  7. Copy and Paste Features
    1. SSI Web now supports copy and paste between two simultaneously open instances of SSI Web.
    2. Copying and pasting lists from within SSI Web and from outside SSI Web are now handled in the same way, so the "Paste from Word" button is no longer needed.
    3. Lists of items may now be copied and pasted out of SSI Web and into other packages.
    4. List index numbers in front of lists pasted from Word are now stripped when copied into SSI Web.
  8. Don't Know/Refused code in Numeric Questions - A Don't Know/Refused code is now allowed in Numeric questions.
  9. Export and Import of Skips and Quotas - Skip Logic and Quota Logic may be exported and imported from *.csv files.
  10. New SSI Script Functions
    1. ToolTip( ) - ToolTip lets you more easily insert a Tooltip in SSI Web. A Tooltip is associated with either text or graphics and displays a pop-up message when respondents mouseover the text or object.
    2. ProgressBarSet( ) - ProgressBarSet lets you override and set the progress bar for any page in the survey.
    3. SetValue( ) and GetValue( ) - SetValue and GetValue let you save new variables equal to respondent answers or other scripting logic, and retrieve (get) those values in later pages of the questionnaire.
  11. Progress Bar Update - The progress bar area has been updated to include the option of tracking progress based on how many questions have been seen out of the total number of questions. You additionally can specify what % of the total number of questions you expect each respondent will see. This is helpful so that the progress bar gives more consistent progress reports to respondents if you are doing randomized blocks or looping.
  12. Internal List Labels - Lists now support abbreviated labels with each list element, for use in data labels and other data analysis areas. This is especially helpful if your list elements include Javascript code or HTML syntax that you don't want to be used in the data labels during the reporting phases.
  13. Mootools to jQuery - The Javascript library used by SSI Web has changed from Mootools to the more industry-standard jQuery.

The following features have been added to CBC:

  1. CBC attribute size - The number of attributes in the Advanced Design Module has increased to 250. Alternative-specific designs and other interesting advanced applications (such as Menu-Based Choice studies) can benefit from having an experimental designer that can handle up to 250 attributes.
  2. CBC Conditional Relationships - CBC conditional pricing and conditional display are no longer limited to 4 attributes.
  3. Time per task in .cho file - Time per task is now (optionally) included in CBC export of .cho file. This information is not used in utility calculations, but may be helpful or interesting to the researcher.

The following features have been added to MaxDiff:

  1. Position Report - MaxDiff export now also exports the position numbers of selected items, which can be useful for identifying straightliners and other patterns of responses.
Read More

Version 7 History

17 May 2012

Note that except as noted, these bugs were also fixed in SSI Web 8.0.0.The following feature has been added to ACBC:

  1. When calculating ACBC utilities, analysis constraints left blank are not constrained. For example, entering <2, 1, blank> as the relative preferences of a three-level attribute on the constraints dialog will cause only the 1st and 2nd levels to be constrained. Although this feature has been in the software since ACBC shipped, it has not been documented.

The following bug fixes have been made in SSI Web:

  1. Placing a 0. in a numeric min would translate on the server as [%.%] and would result in a scripting error. This has been fixed. Also, [%.%] will no longer pass study verification as a valid numeric value.
  2. With a radio button on the last page and graphical submit turned off, clicking the "Enter" key while the focus is on the radio button will cause the page to submit twice. The error message "Your survey has already been submitted. Thank you." would be shown. This has been fixed.
  3. When using CAPI 3.0 with studies created by SSI Web 7.0.24 or 7.0.26, an error would be shown while loading the study in CAPI stating a file could not be found. This has been fixed and SSI Web 7.0.28 will create CAPI files that work with CAPI 3.0.
  4. Error highlighting for column-based grids did not always work correctly. This has been fixed.
  5. Auto upload has been improved and should now be faster and have better compatability with web servers.

The following bug fixes have been made in CBC:

  1. Attribute labels in CBC would not always display correctly when Partial Profile and Conditional Relationships were enabled. This has been fixed.
  2. CBC Test Design now works when the design has only one concept.
  3. Adding CBC questions to a questionnaire with existing CBC questions now adds the new questions after the CBC exercise instead of at the end of the questionnaire.

The following bug fixes have been made in ACBC:

  1. If an attribute is specified to not be included in the BYO section in ACBC, and then the respondent answers in such a way that that the attribute is shown in a Must-Have question, and it is selected, the Screener section will end immediately. This has been fixed.
  2. When calculating ACBC utilities, analysis constraints with the value of 0 were being treated as not constrained. For example, entering <0, 1, 2> as the relative preferences of a three-level attribute on the constraints dialog would only cause the 2nd and 3rd levels to be constrained. The same would hold true for constraints customized based on other survey data (i.e. using a question name for the constraint). This bug has been fixed. A notice containing more information sent to ACBC users is available by clicking here. (Note that this bug was not fixed in version 8.0.0 but was fixed in version 8.0.2.)
  3. ACBC Analysis automatically calculates price values for the summed pricing attribute. If level prices were set up such that some levels within an attribute had a component price and some were left blank (not 0), the automatically calculated price values would not interpret this as a 0 and would ignore that level. This has been fixed.
  4. Randomized blocks that contained only an ACBC exercise that exited early from the tournament section would skip directly to the next question regardless of the block or study order. This has been fixed.
  5. ACBC would allow the survey author to put ACBC questions of the same type on the same page. If that happened, the Screener and Tournament questions would show the same designs for the questions on the same page. Must-Have and Unacceptable questions also had this problem. Besides design problems, this might cause incorrect counts and utility calculations. This bug has been fixed and now only one ACBC question is allowed per page. (Note that this bug has not yet been fixed in any version 8 code.)
Read More

29 August 2011

The following bug fixes have been made in SSI Web:

  1. There was a bug in quota cell randomization. The bug could either cause the randomization to always be the same or it could cause respondents to be disqualified who would otherwise be qualified. This bug was introduced in version 7.0.0 and has been fixed.
  2. Using the Opera browser to take a survey in version 7.0.24 would sometimes not allow the respondent to submit the page. This would happen on pages with grid questions that used list items marked as Respondent Specify ("Other Specify"). The submit button would also have to be a text button (not a graphic). In 7.0.24 we updated to the latest MooTools JavaScript library -- we have fixed this bug by rolling back to the version that was shipping prior to 7.0.24.
  3. A lab license would not allow editing quota limits in the Admin Module. This has been fixed.

The following bug fix has been made in ACBC:

  1. Having an ACBC exercise in a randomized block could cause the randomized blocks to function improperly. For this to occur, the exercise must have tournament questions but not have calibration questions, and the question following the ACBC exercise must not be in a randomized block. Additionally, it would only happen when part of the tournament section was not shown (such as when tournament questions are dropped because of answers to prior questions). Under these conditions, the question following the randomized block was always loaded and shown after the ACBC exercise. This has been fixed.

The following bug fix has been made in MaxDiff:

  1. JavaScript verification for unique answers would fail with custom MaxDiff questions implemented with Free Format. This has been fixed.

The following change has been made in SSI Web:

  1. The MooTools JavaScript library has been rolled back from version 1.3.2 to 1.2.1.
Read More

05 July 2011

The following bug fixes have been made in SSI Web:

  1. Printing a study to PDF would give the error "HTML render is blank" if Internet Explorer 9 was installed on the machine. This has been fixed.
  2. The SSI Script function ADDSORTED() resulted in different ascending or descending order each time when breaking ties. This has been fixed.
  3. If a password field matched the converted username or password in a study authored with SSI Web version 6 or earlier, then exported username or password data would be blank. This was introduced in version 7.0.16 and has been fixed.
  4. The SSI Script function TextExists() was incorrectly treating the text to search for as a Perl regular expression. This means for example that: TextExists(Q1, ".*") would match anything. Certain characters such as ".", "*", "(", "$", etc. had special meaning and could have caused TextExists() to not work properly. This has been fixed.
  5. On many questions, Cell Padding and Body Border could not be set to 0. This has been fixed.
  6. Grid cells with labels would always be left aligned in Chrome and Safari. This has been fixed.
  7. If a respondent uses the long URL (the URL that SSI Web creates on each page) to re-enter the survey, the same browser and IP address that began the survey are now required. Respondents can still restart surveys using the login page (or one click link) with a different browser or IP address.
  8. Auto Upload account names were not required to be unique. If two account names had the same name and the second account was selected when the dialog was closed, the first account would be selected when the dialog box was reopened. This has been fixed.
  9. A grid variable that used ranking combo boxes or select combo boxes that had "Randomize Response Options" checked (before being changed to a constructed list) would randomize the constructed list. This was in addition to any randomization the constructed list might have had. The additional randomization was not recorded. This has been fixed and the “Randomize Response Options” checkbox is ignored with constructed lists.

The following bug fix has been made in ACA:

  1. When backing up to an ACA calibration question using horizontal layout, only the first calibration answer was restored. The remaining answers were empty.

The following bug fixes have been made in CVA:

  1. During CVA HB, Monotone, and OLS analysis, a CVA task response with a decimal value (e.g. 4.2) was not being properly checked to make sure it fell within task minimum and maximum response values. This has been fixed.
  2. Filtering tasks in CVA OLS gave the following error during computation: Object doesn't support this property or method. This was introduced in version 7.0.0 and has been fixed.

The following change has been made in SSI Web:

  1. The MooTools JavaScript library has been updated from version 1.2.1 to 1.3.2.
Read More

09 February 2011

The following bug fix has been made in SSI Web:

  1. A bug in passwords would cause, under certain extremely rare conditions, incorrect data to be written to a respondent's merged fields. The bug could also permit (also under extremely rare conditions) a login that is not in the password table, or deny a valid login. This has been fixed. We view this as a critical update and encourage all version 7 users to upgrade to this version. Please contact Sawtooth Software for more information.
Read More

26 January 2011

The following bug fix has been made in SSI Web:

  1. Importing a translated study (from an XLIFF file) was removing newline characters from the text. This has been fixed.

The following bug fix has been made in ACA:

  1. The formatting settings of the Scale Header in ACA Rating and ACA Importance questions did not work. This bug was introduced in version 7.0.0 and has been fixed.

The following bug fix has been made in ACBC:

  1. In some cases involving a constructed attribute list and a constructed level list, a bug would still exist that we described in an earlier bug fix (see version 6.6.4 ACBC bug fixes, item #2 -- it was to prevent including extra information in the CHO and HB build for missing levels when no levels were missing.) This has been fixed. Additionally, using constructed attribute lists would also sometimes cause an error to be shown during analysis that a key was missing in the data file. This has been fixed.

The following bug fix has been made in CBC:

  1. Print Study to a text file would give an error if printing a CBC exercise that included Null Prohibitions. This bug was introduced in version 7.0.0 and has been fixed.
Read More

13 December 2010

The following bug fixes have been made in SSI Web:

  1. The Chrome browser would not detect if JavaScript was enabled properly. This would only happen when linking directly into the survey (bypassing the login.html page) using the "One-Click" login method. As a result both JavaScript and Perl verification would fail on the first page seen by the respondent. Also, JavaScript was recorded as not enabled (or 0) in the data file. All pages after the first page seen would work properly.
  2. The "Disable Browser Back Button" feature would make it impossible to submit a page using the Safari browser. SSI Web now allows a respondent using Safari to go forward and ignores the "Disable Browser Back Button" feature.

The following bug fix has been made in ACA:

  1. In ACA exercises, if "Randomize Attribute Order in Rating and Importance Questions" was turned on and the respondent used the Previous Button to back up to a random ACA Rating or ACA Importance, then it was possible for a some ACA questions to be skipped and not recorded in the data file. This has been fixed.

The following bug fix has been made in MaxDiff:

  1. MaxDiff "Ask Best Only" Style 1 (radio buttons on left side of labels) would change to Style 2 (radio buttons on the right side of the labels) if a MaxDiff question was edited. The reverse would also happen -- Style 2 would change to Style 1. This bug was introduced in version 7.0.16. It has been fixed.
Read More

04 November 2010

The following bug fix has been made in SSI Web:

  1. Double quotes in error messages would cause the web page to not submit during an interview. This bug was introduced in version 7.0.16. It has been fixed.
Read More

02 November 2010

The following feature has been added to SSI Web:

  1. Auto Upload has now has the ability to change between active/passive modes. SSL/TLS support was also added.

The following bug fixes have been made in SSI Web:

  1. Post skipping to a question that auto completes when that page is not shown would only save constructed list data. The question data was missing. This has been fixed.
  2. Placing custom SSI Script or Unverified Perl in Error Messages would sometimes cause the error messages to not work correctly and prevent the survey page from being submitted. This has been fixed.
  3. When randomizing questions on a page, questions before the anchor question were not shown when using the back button. This has been fixed.
  4. When restarting a survey in the middle of a randomized block set, the next block or next page would be shown instead of the restart question. This has been fixed.
  5. If multiple groups of randomized questions were on the same page, the randomized order data were saved as one group instead of multiple. Additionally, the data were saved as if the first question on the page was the anchor question. This could result in the data not being exportable from SSI Web. This has been fixed and the randomized order data are saved as separate groups assigned to the correct anchor questions.
  6. The Export Text for Translation feature was not exporting conditional text or prices for CBC, CVA, or ACBC. This has been fixed.
  7. When importing translated text, whitespace (e.g. spaces) at the end of the original text would show the Import Language Conflict dialog box. This has been fixed.
  8. If the study contained both text and numeric data types in Password, Merged, or Pass-In fields, and the text variable data was longer than 15 characters, an export error would be shown that the width was too large. This has been fixed.
  9. Auto upload was not adding files to the recovery backup folder. This has been fixed.
  10. Converting a study created with a version prior to 6.2 would export blank password and username data. This bug was introduced in version 7.0.0. It has been fixed.

The following bug fix has been made in CBC:

  1. Calling the SSI Script function CBCDesignLevelText() inside of a CBC question would cause the survey to hang for the respondent. This has been fixed.

The following bug fix has been made in CVA:

  1. When calculating CVA utilities using CVA/HB, Ordinary Least Squares, or Monotone Regression, the build would report an incorrect number of constraints when constraining "to the a priori rankings pre-specified on the Attributes dialog". Although the report is incorrect, the correct constraints are used in analysis.

The following bug fix has been made in MaxDiff:

  1. MaxDiff "Ask Best Only" Style 2 (radio buttons on the right side of the labels) would display Style 1 (radio buttons on left side of labels). This has been fixed.
Read More

20 August 2010

The following bug fix has been made in ACBC:

  1. A bug was fixed in ACBC that had to do with not including attributes in the BYO. In 7.0.10 we introduced the ability to not include attributes in the BYO with a preference order set to "None". In this situation, we choose a random level when recording the BYO answer. If more than 1 of these attributes was not included, it was possible that a random level could be chosen that was greater than the possible number of levels for that attribute. This has been fixed.

The following bug fix has been made in MaxDiff:

  1. If a constructed list was used in a MaxDiff study, the following functions produced incorrect results: MaxDiffRankAttLabel(ExerciseName, Rank#), MaxDiffRankAttValue(ExerciseName, Rank#), and MaxDiffScore(ExerciseName, ItemIndex). This has been fixed.
Read More

12 August 2010

The following bug fixes have been made in SSI Web:

  1. When a Grid question appeared on the first page, the login.html page was not being created for upload to the web. This has been fixed.
  2. In the Admin Module, performing "Search Data" for text containing the @, $, or \ characters would give an error. This has been fixed.
  3. Specifying a post skip to a specific question in a randomized block would cause the skip to go to the specified question instead of the correct random block being loaded. This has been fixed.
  4. When editing a grid question, renaming the question then editing the list corresponding to the question direction (for example, editing the row list of a grid question with a row question direction) would cause the associated list member grid settings to track the list item incorrectly. The settings may have been lost, the program may have given a subscript out of range error, and incorrect or blank input options may have appeared when the respondent took the survey. This would happen when the rename was followed by moving, adding, deleting, cutting, copying, and pasting list items, etc. This bug was introduced in version 7.0.0. This bug has been fixed.
  5. Questions with certain names would cause question verification in the survey to fail. The names that caused failure include: constructor, each, has, keyOf, hasValue, extend, combine, erase, get, set, empty, include, map, filter, every, some, getClean, getKeys, getValues, getLength, and toQueryString. This has been fixed.
  6. Page times were sometimes incorrect by an enormous amount. This has been fixed. Page times are now always calculated in perl for better reliability. This means page times will now include transmission times from and back to the server.
  7. While authoring a study, the width and height of the of an "Other Specify" list member would not load the correct values. These values were being lost or changed. This has been fixed.
  8. While taking a survey, it was sometimes possible to see the quota question on the screen (depending on formatting). This has been fixed.
  9. With quota control, selecting "Check for cell membership" as random with a quota limit of 0 for any cell would cause a server error. This has been fixed.

The following bug fixes have been made in ACA:

  1. In ACA, selecting Skip Importance Questions and Set Prior Importances Based on other Questions would cause the server during data collection to not use the data from the prior questions. Instead, all importances would be set to 3.5. An error log was created and displayed in the Admin Module. This bug was introduced in version 7.0.0. This has been fixed.
  2. In ACA, while editing attributes, if attributes were multi-selected and either moved down or deleted, the questions specified in the "Prior Importances Based on other Questions" may have become associated with the wrong attributes. (This setting is available only when you choose to Skip Importance Questions). This has been fixed.

The following bug fix has been made in ACBC:

  1. Previewing ACBC Unacceptables questions would not work if the attribute had a preference order. This has been fixed.

The following bug fix has been made in CBC:

  1. If level borders in the CBC settings were turned on, the border around the bottom radio button row would not show up. This has been fixed.

The following bug fix has been made in MaxDiff:

  1. An error would occur on the server when using a MaxDiff question in a constructed list command -- for example, AIC(MXD_1_b). This has been fixed.

The following change has been made in ACBC:

  1. ACBC now permits 10 attributes in demonstration mode. Previously only 6 attributes were allowed.
Read More

28 May 2010

The following bug fixes have been made in SSI Web:

  1. Skipping to "End of Block" in a post skip did not work. This has been fixed.
  2. Randomized blocks that used constructed lists that dropped blocks in the set would go to the end of the survey after the last block had been shown. This has been fixed.
  3. Clicking the Back button to a page that had multiple questions where some of the questions were post-skipped over would show all the questions on the page. This has been fixed.
  4. Auto upload incorrectly handled parentheses in path names when uploading, causing the upload to fail. Auto upload would also fail when attempting to upload a file whose name that was longer than 256 characters. These have been fixed. Auto upload also allowed spaces, which are not allowed anymore.
  5. With a grid select question where the rows and columns were both constructed lists that contained only one item, the grid question would be auto answered but would not save the answer in the data file. This has been fixed.
  6. Username/Passwords that were numeric and shorter than 5 characters on export would cause SPSS export to fail and give error code 20. This has been fixed.
  7. The constructed list command "AddSorted" would include items that were empty or had 0 for answers. Now these items will not be included in the constructed list.
  8. "Other Specify" text in Firefox and Chrome displayed on the right of the input box in Ranking and Constant Sum questions. This has been fixed.
  9. Print study would fail if a ranking question was printed graphically with "Show Variable Names" turned on. This has been fixed.
  10. HTML was being removed in error messages. This has been fixed. HTML is now allowed in error message text.

The following bug fixes have been made in ACBC:

  1. Attributes that were removed from the BYO that were also invovled in a prohibition were not prohibited correctly in the first calibration concept. This has been fixed.
  2. In ACBCBYO if an attribute was involved in a prohibition but was not displayed because it was dropped due to a constructed list, the ACBCBYO question would not submit. This has been fixed.
  3. An ACBC survey with a constructed attribute list where some attributes were not shown in the BYO would sometimes show an error during the build portion of ACBC analysis. The error stated that the respondent was skipped because there were no eligible tasks for estimation. Instead of showing an error, the respondent should simply not have been included in the analysis. This has been fixed.

The following bug fix has been made in CBC:

  1. SSI Script inside of CBC attribute text returned by the SSI Script function CBCDesignLevelText() would not evaluate. For example, custom CBC tasks would not display graphics properly if using [%GraphicsPath()%]. This has been fixed.

The following bug fix has been made in MaxDiff:

  1. SSI Script functions "MaxDiffRankAttValue" and "MaxDiffRankAttLabel" were using a random number generator which would not sort the same way on ties every time. This caused constructed lists to sometimes see duplicate items which were then removed, resulting in an incorrectly created ranking. This has been fixed.

The following bug fix has been made in CBC, CVA, and MaxDiff:

  1. Paper and Pencil task formatting was being applied to only the first task. The same bug existed in Print Study to HTML. This has been fixed.

The following change has been made in ACBC:

  1. In ACBC, specifying an attribute to be excluded from the BYO question no longer requires that a preference order is specified for the attribute.

The following change has been made in CBC:

  1. The academic license now supports Advanced Design Module (ADM) features.
Read More

22 March 2010

The following bug fixes have been made in SSI Web:

  1. Free Format 'text' variables where the survey author included the 'value' attribute for initial values in the HTML would return bad HTML if respondents answered the question then used the "previous" button to return to the question. This has been fixed.
  2. Page times would randomly produce enormous times separated by a comma. This has been fixed.
  3. Grids with multi-select would export incorrectly when using either of the 'Item Numbers' data formatting export options. This bug was introduced in version 7.0.0. and has been fixed.
  4. In the Numeric and Open End question HTML there were multiple "id" attributes with the same name. This bug was introduced in version 7.0.0 and has been fixed. The "id" for the whole question is now "id=QuestionName_div".
  5. Auto Upload was setting the permissions for cgi-bin such that the study would not work on some servers. This has been fixed.
  6. Graphical submit buttons did not convert correctly from a version 6 study to a version 7 study. This caused the graphical submit button not to display in Internet Explorer. Now the graphical submit HTML will not be converted to version 7.
  7. Older versions of Firefox (i.e. version 2.0) would not display the graphical next button. This has been fixed.

The following bug fixes have been made in ACBC:

  1. ACBC exercises that had no price attribute, but did have either prohibitions, conditional text, or conditional price, would display an error message during the interview by mistake. The error said that the price attribute was missing. This bug was introduced in version 7.0.0 and has been fixed.
  2. Formatting options for the body text in ACBC Must Have and Unacceptable question types did not work. This bug was introduced in version 7.0.0 and has been fixed.

The following bug fix has been made in ACA:

  1. Input Box Justification for the ACA Calibration question did not save if right justification was selected. This has been fixed.

The following bug fixes have been made in CBC:

  1. Using CBC Shelf Facing with the none option displayed in the task header or footer with [%CBCNone()%] did not work correctly with graphical select buttons. This bug was introduced in version 7.0.0 and has been fixed.
  2. When using CBC Shelf Facing where the none option was displayed with [%CBCNone()%], the "previous" button would not redisplay the prior answer correctly if the none option was selected. This has been fixed.
  3. Attribute label formatting in CBC did not work correctly. This bug was introduced in version 7.0.0 and has been fixed.

The following change has been made in CBC, CVA, and MaxDiff:

  1. Respondents accumulated via paper-and-pencil will now always be marked as "complete". Previously, respondents with any missing answers were marked as "incomplete".

The following features have been added to SSI Web:

  1. When exporting, .tab and .sav files may now be exported with either dynamic or fixed widths.
  2. When exporting, the .labels file exported with a .csv is now optional. Not exporting the .labels file will improve speed.
Read More

26 February 2010

The following bug fixes have been made in SSI Web:

  1. If a constructed list was used in a question and reused in another question where Auto Advance was triggered (because the constructed list had only one item), the data for all questions in between these questions including the second question to use the constructed list were deleted from the data file (just for that respondent). Auto Advance only triggers for Select, Ranking, Constant Sum, and Grid question types. This bug was introduced in SSI Web 7.0.0. It has been fixed.
  2. If a question was auto-advanced, the answer and the created constructed list for that question were not available to subsequent questions until after the next question was shown to the respondent. For instance, if a question used a constructed list and was auto-advanced, and the next question used the first question or its constructed list in its constructed list commands, that next question's constructed lists would not contain any information from the first question or its constructed list. This has been fixed.
  3. If a question was auto-advanced, post skips associated with that question were not executed. This has been fixed.
  4. The export option "Export multiple-response questions as item numbers (single column of chosen items)" would always export blank data. This has been fixed.

The following change has been made in SSI Web:

  1. When using the SSI Web Back button, free format hidden variables were not pre-filled. This was originally by design so that backing up would not change data that respondents could not see. However, this caused some data to give unexpected results since JavaScript would not re-run to change the hidden values. This has been changed so free format hidden variables are now prefilled.
Read More

18 February 2010

The following bug fixes have been made in SSI Web:

  1. In version 7.0.0, the Perl files were in PC file format, which caused them to not work on some servers. This has been fixed.
  2. Auto Advance was answering questions that used a predefined list with only one item in the list. An additional error would occur if the question was on the start page. This has been fixed.
  3. "None of the above" in Grid Questions did not uncheck all other checkboxes. This has been fixed.
  4. Turning off "Require Response" on Open End Questions and Free Format (Text and Text Area) Variables did not function properly. During interviewing, the "Minimum Number of Characters" was still enforced even if "Require Response" was unchecked. This bug was introduced in version 7.0.0 and has been fixed.
  5. Randomized Blocks would always drop the last block if the last block did not come immediately after another block in the same set. This has been fixed.
  6. An incorrect randomized block would be shown if previous blocks of a set had been skipped over and not shown, and it was the first time encountering any block of that set. This has been fixed.
  7. Graphical Next would not function at all on the start page if JavaScript was disabled.
  8. Password reports did not work. This has been fixed.
  9. ACBC BYO Levels did not respond to formatting changes. This has been fixed.

The following enhancement has been made in SSI Web:

  1. When accumulating paper-and-pencil data, errors in the file to accumulate are now shown together instead of one respondent at a time.
Read More

26 January 2010

The following features have been added to SSI Web:

  1. Study names may now be up to 24 characters. No more limitation of 4 character study names.
  2. Multiple CBC studies, ACA studies, and CVA studies may be included within the same project. This means less linking of surveys!
  3. Back button (not just the back button on the browser toolbar, but a back button within the survey). You can even back up to all parts of the survey after a restart.
  4. Graphical select and radio buttons (which also means you can dramatically change the size of buttons). Also, an option to highlight the text of response items on mouseover.
  5. Randomized Blocks of questions (e.g. the 3 blocks of questions: Q1-Q3; Q4-Q10; Q11-Q25 can be randomized, and Q1-Q3 always stay together as a block, Q4-Q10 always stay together, etc.)
  6. The standard formatting controls in the edit question dialogs will operate nicely in conjunction with styles. In previous versions, styles typically would override most of the formatting controls. Starting in v7, SSI Web surveys by default use a more attractive style, so your surveys will look much better even before you apply your favorite style.
  7. Many new styles available in the library.
  8. Dramatically improved data export speed.
  9. Ability to export all the text for your survey to an XML XLIFF file format (commonly used by translation houses). Once your survey is translated, you can import the XLIFF file back into your survey in one step.
  10. Automatic survey upload and data download. The FTP process is built into SSI Web, so this feature will make working with the server faster and less complicated. SFTP (a more secure file transfer protocol) is also supported.
  11. Ability to edit data (change responses) through the online administrative module, including changing the status of a respondent from an incomplete to a completed record.
  12. Ability to set filters for the marginals report. This lets you, for example, tally the responses for a question among respondents belonging to large businesses.
  13. Incompletes (drop out) report. Reports what percent of respondents dropped out at each question.
  14. Internal respondent numbers are now all numeric, not alpha characters. Numbers are much easier to interpret and work with!
  15. Ability to save response time (latency) per page. This is on by default, but you can easily turn it off (to reduce the size of the stored data file) via SSI Web's interface on the Survey Settings dialog, Advanced tab.
  16. We've redone the way that "error messages" work (messages that tell respondents that they've missed a question on the page, or used an invalid response). In v7, we highlight the question with the problem using a (typically red) box, and place a (typically red) message within the box describing the nature of the error. (You can change the warning color from red.)
  17. Auto advance for questions involving constructed lists that include a single response item with a required response. Previously, the respondent would see the question and be forced to check the one (and only) possible response. With v7, we recognize that the response is a foregone conclusion, and we automatically record the answer and advance to the next question.

The following features have been added to MaxDiff:

  1. There is a new function for computing MaxDiff scores on-the-fly using individual-level logit. You can return scores for each item, or a ranking of the items. This makes it possible, for example, to force just the top (most important) n items onto a new constructed list.
  2. Paper-and-Pencil data collection and accumulation methodology is now incorporated within SSI Web. No more need to use the separate MaxDiff Designer software.
Read More

Version 6 History

10 February 2011

The following bug fix has been made in SSI Web:

  1. A bug in passwords would cause, under certain extremely rare conditions, incorrect data to be written to a respondent's merged fields. The bug could also permit (also under extremely rare conditions) a login that is not in the password table, or deny a valid login. This has been fixed. We view this as a critical update and encourage all version 6.2 and later users to upgrade to this version. Please contact Sawtooth Software for more information.
Read More

18 January 2011

The following bug fixes have been made in SSI Web:

  1. An error may be shown when running ACBC, CBC, ACA, CVA, or MaxDiff analysis on a study that contains non-ASCII (international) characters. The error includes the text "Can't locate utf8.pm in @INC". This bug was introduced in version 6.6.14 and has been fixed.
Read More

07 January 2011

The following bug fixes have been made to ACBC in SSI Web:

  1. In some cases involving a constructed attribute list and a constructed level list, a bug would still exist that we described in an earlier bug fix (see version 6.6.4 ACBC bug fixes, item #2 -- it was to prevent including extra information in the CHO and HB build for missing levels when no levels were missing.) This has been fixed. Additionally, using constructed attribute lists would also sometimes cause an error to be shown during analysis that a key was missing in the data file. This has been fixed. Note: This bug was also fixed in SSI Web 7.0.21.
  2. Previewing ACBC Unacceptables questions would not work if the attribute had a preference order. This has been fixed. Note: This bug was also fixed in SSI Web 7.0.12.

The following bug fixes have been made to ACA in SSI Web:

  1. In ACA, while editing attributes, if attributes were multi-selected and either moved down or deleted, the questions specified in the "Prior Importances Based on other Questions" may have become associated with the wrong attributes. (This setting is available only when you choose to Skip Importance Questions). This has been fixed. Note: This bug was also fixed in SSI Web 7.0.12.

The following bug fixes have been made to SSI Web:

  1. The Chrome browser would not detect if JavaScript was enabled properly. This would only happen when linking directly into the survey (bypassing the login.html page) using the "One-Click" login method. As a result both JavaScript and Perl verification would fail on the first page seen by the respondent. Also, JavaScript was recorded as not enabled (or 0) in the data file. All pages after the first page seen would work properly. Note: This bug was also fixed in SSI Web 7.0.20.
  2. The "Disable Browser Back Button" feature would make it impossible to submit a page using the Safari browser. SSI Web now allows a respondent using Safari to go forward and ignores the "Disable Browser Back Button" feature. Note: This bug was also fixed in SSI Web 7.0.20.
  3. Radio buttons would not vertically align properly in Internet Explorer. This has been fixed.
Read More

22 March 2010

The following bug fixes have been made to SSI Web:

  1. When accumulating CBC or CVA paper-and-pencil data, the following error is given: "Internal Respondent Number 'x' is missing the key 'sys_PasswordChecksum'." This bug was introduced in SSI Web 6.6.10 and has been fixed.
Read More

12 February 2010

Note: These bugs were also fixed in SSI Web 7.0.0The following bug fixes have been made to ACBC in SSI Web:

  1. A problem existed in the Unacceptables Section of ACBC Counts. Once the data file encountered a respondent that had marked price as unacceptable, all following respondents that did not have a price marked as unacceptable would have a 0 in this column instead of a blank. This caused the reported average unacceptable price to go down drastically (the base was incorrect). This bug has been in the software since the introduction of ACBC (version 6.6.0).
  2. SSI Web would crash if text was entered into the BYO left panel width or right panel width fields during authoring. This has been fixed.

The following bug fixes have been made to ACA in SSI Web:

  1. With certain settings, SSI Web could crash when changing ACA study parameters. This bug has been fixed.

The following bug fixes have been made to SSI Web:

  1. During an interview, if random quota assignment was turned on, two quota questions with the same number of cells would have the same randomization for a respondent. This has been fixed.
  2. When converting a study from SSI Web 6.2 to SSI Web 6.4 or 6.6, if a question or MaxDiff exercise named "Password" existed it would incorrectly be renamed to "Start". Converting 6.2 to version 6.6.10 will not cause this problem.
  3. When using the HTML tag or Custom JavaScript verification on the "Start" question, these areas were not included when the software checked for errors in the questionnaire. If those areas contained SSI Script with errors, they were not being reported. This has been fixed.
Read More

28 October 2009

The following bug fixes have been made to ACBC in SSI Web:

  1. There was an error in ACBC Price Adjustments. The problem occurs if you have Price Adjustments defined for an attribute and that attribute has one or more level prices left blank. If this occurs, during data collection the total price for the BYO could be saved incorrectly in the data file. Please read this important bug notice. This bug has been fixed.
  2. There is potential for error in ACBC Summed Pricing. When entering the "List of Levels" on the "Attributes" tab in ACBC, copying and pasting the List of Levels instead of choosing it from the combo box causes pricing information to be incorrect for that attribute. If the survey author does not visit the Pricing tab and view pricing for the attribute (even if the price is 0 or blank for each level), incorrect pricing information will be recorded during data collection. If you believe you have been affected by this bug, please contact Sawtooth Software.
    Additionally, copying and pasting the "Include in BYO" setting on the Attributes tab could cause this change to not be reflected in the analysis and cho export settings.
    These copy and paste bugs have been fixed.
  3. This bug only affects ACBC "None" Threshold Estimation where HB or Monotone Regression analysis was done on a computer that uses a comma to represent the decimal value. Decimal values were written to the .hbucal file using commas instead of periods. This has been fixed.

The following bug fixes have been made to ACA in SSI Web:

  1. If the default ACA calibration scale was changed (no longer 0 to 100), the ACD export in ACA studies was not scaling calibration responses back to the 0 to 100 scale. Instead, the actual responses were being written to the ACD. A file calibrated in ACA/HB would be scaled as if the responses were on the 0 to 100 scale, resulting in incorrect utilities. This bug has been fixed.

The following bug fixes have been made to CBC and CVA in SSI Web:

  1. This bug only affects the use of the Advanced Prohibitions tab (not the standard Prohibited Pairs tab) in CBC and CVA. (Advanced prohibitions are only allowed within the Advanced Design Module of the CBC product. Therefore, CBC users without the Advanced Design Module are unaffected.) The bug is triggered if you enter the Prohibitions Dialog "Advanced Prohibitions" tab. If a design exists and you change prohibitions within the Advanced Prohibitions tab and immediately click the OK button (without switching to another tab), the CBC or CVA design is not deleted. A user could then take the study to field with the outdated design that does not include the prohibition changes made on the advanced tab. Additionally, if the user does regenerate the design, a message is given saying "It appears that a current CBC design file already exists for this exercise", even though the design is not current. This bug was introduced in version 6.6.0. This bug has been fixed.
  2. The CVA design file was not being put in the recovery folder. This has been fixed.

The following bug fixes have been made to SSI Web:

  1. The SSI Script key words "AND", "OR", "NOT", and "MOD" were allowed as names for questions, lists, passwords, merged variables, pass-in variables, and quota cells. Using one of these key words as a name described above and then using it (either as a key word or a name) in SSI Script would give undefined results. In the best case, the logic would be flagged during study verify. In the worst case, unintended logic could have been executed on the server during the interview which could affect the survey results. This has been fixed.
  2. Changing the name of a parent list would cause a disassociation with previously associated constructed lists. Study Verify would catch this before going to field. This bug was introduced in version 6.2.0. This bug has been fixed.
  3. Changing the case of the name of a parent list of a constructed list would cause a crash while authoring the survey. This bug was introduced in version 6.2.0. This bug has been fixed.

This release also includes some security updates for the server code used during data collection.

Read More

30 July 2009

The following bug fixes have been made to ACBC in SSI Web:

  1. Otter’s method did not work properly with constrained estimation. This bug was introduced in version 6.6.4. This has been fixed.
  2. Constrained estimation for piece-wise Summed Price was not stable, with parameters for Summed Price growing huge and overall model fit shrinking to zero. This bug was introduced in version 6.6.4. This has been fixed.
  3. Constrained estimation for attributes specified as linear coefficients (rather than the default part-worth function) could sometimes be constrained in the opposite direction (i.e. positive instead of negative) as defined by the researcher. If this affected you, it would be painfully obvious, as the direction of the coefficient would be opposite what had been requested, and the overall model fit degraded. This bug was introduced in version 6.6.0, the initial release of ACBC. This has been fixed.

The following bug fixes have been made to CBC and CVA in SSI Web:

  1. Conditional Relationships were not responding correctly to the button click to move levels up in the list of levels. Normally, moving levels of an attribute will cause information found in rows in conditional relationship tables to follow the movement of the associated level. With this bug, the conditional relationship tables were not responding to the level being moved, or they were responding by moving data rows of the levels of the wrong attribute. This bug was introduced in version 6.6.0. It affects both CBC and CVA, but not ACBC. It has been fixed.
Read More

21 July 2009

The following bug fixes have been made to ACBC in SSI Web:

  1. We found a weakness in our seeding for random numbers when generating the ACBC experimental design. As far as we know, this is a bug that is quite obscure, and only affects smaller-than-recommended studies, where just a very few attributes were included in BYO. The bug could cause multiple attributes to receive the same randomization seed, leading to correlated designs across attributes. This has been fixed.
  2. We discovered that when constructed lists are used just to randomize ACBC attributes (without dropping any levels), we were including some unneeded information in the .CHO files and in the build of the design matrix for HB estimation. We were including an extra uninformative column (independent variable) and uninformative additional BYO choice tasks (those comparing threshold parameters to included vs. excluded levels). HB is robust enough that the additional uninformative information was not causing any practical damage to parameter estimation. But, we've fixed the error, and the information used in the design matrix is now cleaner and more proper.
  3. When calculating ACBC utilities using HB, if writing out .dra files and using constraints, unconstrained draws were being written to the file. We've fixed that, and now constrained draws are written.
  4. When calculating ACBC utilities using HB, the RLH (fit) value was being improperly written to .dra files. Values were becoming either 0 or 1, because we were forgetting to multiply by 1000 before rounding to the nearest integer. This has been fixed.

The following changes have been made to ACBC in SSI Web:

  1. We recently have come to the opinion that the ACBC Unacceptables questions should take precedence over ACBC Must-Haves. So, the default settings in the software will place the Unacceptables first, followed by Must-Haves. Just switch the previous decision rules we trained you on regarding how many Must-Haves and Unacceptables to include in the questionnaire, and their positioning. The new software will build questionnaires with Unacceptables above Must-Haves by default (you can still rearrange them, if you wish, to give Must-Haves precedence as before).
  2. This version of ACBC allows greater flexibility regarding the minimum and maximum number of attributes you can change from the BYO-selected levels when generating near-neighbor concepts. Some users expressed that the previous settings didn't give them enough flexibility to do what they wanted to do. And, we recently found for a very small ACBC study (that only included 2 attributes in the BYO section, and 2 attributes that were skipped in BYO, for a total of 4 attributes in the study), that we needed flexibility to sometimes vary 0 attributes from BYO levels (previous minimum possible was 1), and to vary more than 1/2 of the attributes (previous limit was no more than half). The new limit is 1/2 the number of attributes in BYO +1 (rounded down).
  3. In the first release of ACBC, we used a new streamlined strategy for constrained utility estimation in HB that did not involve maintaining both constrained and unconstrained betas for each respondent. With a subset of the attributes being constrained, this seemed to work fine in our previous testing and projects. But, we recently did a fully constrained model (all attributes) for a project including some globally-constrained attributes and some customized-constrained attributes (based on ratings questions that each respondent answered). The streamlined constraints algorithm was not stable when all attributes were being constrained, with elements in the covariance matrix running off to infinity (and the part-worth parameters also becoming overly huge). We have reverted to the method as used in CBC/HB (simultaneous tying, maintaining both unconstrained and constrained part-worths in the MCMC process), and this resolves the problem.
Read More

19 March 2009

The following bug fixes have been made to SSI Web:

  1. This bug only affects the use of the Advanced Prohibitions tab (not the standard Prohibited Pairs tab) in CBC and CVA in SSI Web version 6.6.0.  (Advanced prohibitions are only allowed within the Advanced Design Module of the CBC product.  Therefore, CBC users without the Advanced Design Module are unaffected.)  The bug is triggered if you click Compose + Conjoint Settings + Prohibited Pairs... and then click the Advanced Prohibitions tab.  Then if you specify prohibitions within the Advanced Prohibitions tab and immediately click the OK button, the prohibitions you just specified are not actually saved and used in generating the design.  This bug has been fixed.
Read More

27 February 2009

Adaptive Choice-Based Conjoint (ACBC) is now available in the SSI Web platform.  For information on ACBC including purchase information, please visit the ACBC product page on our website.

The following enhancements have been made to SSI Web:

  1. We've added a new sys_skipto option that can be used in the URL when linking to a study.  This new option allows you to restart a survey at a specific question.  This feature really helps if you are linking multiple studies and you are using Quota Control.  You can now link out of one survey to another survey and then return again to finish the original survey.
  2. New SSI Script functions have been added: RIC (Remove If Chosen), DisplayListLabels, FormatListLabels, ProgressBarOff, Studyname, FormatNumber, RoundToNumber, CBCAttributeLabel.

The following bug fixes have been made to SSI Web:

  1. The Disable Back Button feature did not work in Firefox.  This has been fixed.
  2. This bug only affects analysis done on a computer that uses a comma to represent the decimal value.  Entering a decimal value in MaxDiff Scores Prior Variance, CVA HB Prior Variance, CVA Analysis Scale Min and Max, and other analysis settings that accept input with decimal points could have the value changed when returning to the analysis settings dialog where the value is displayed.  For instance, entering 1.5 for the prior variance on a German language machine will have a blank value for prior variance when you return to the settings dialog.  (Analysis done before returning to the dialog will use the correct value, 1.5 in the example.)  Our testing has shown the value will be blanked out and that the user will be prompted to re-enter a value.  However, we can not rule out that the value may be different and the user will not be prompted to re-enter a value, causing the changed value to be used in analysis.  This has been fixed.
Read More

31 October 2008

The following features have been added to SSI Web:

  1. You can now export multiple select (checkbox type questions) as a single column of item numbers that were chosen

The following changes have been made to SSI Web:

  1. On November 1, 2008 Sawtooth Software began using a new license agreement for its products.  Please read the new license agreement (available in the SSI Web install folder and during installation).
  2. All conjoint products (when installed without a CiW license) now have 50 CiW data fields available. Previous versions only had 10.
  3. The CiW 50 license now allows 100 CiW data fields.

The following bug fixes have been made to SSI Web:

  1. If a quota cell limit was set to 0 then the Admin Module would display the same graph as the prior quota cell graph.  This has been fixed.
Read More

21 July 2008

The following bug fixes have been made to SSI Web:

  1. A problem was fixed related to constructed lists and quota logic.  When a constructed list is created it is saved in memory and then to disk at a later point.  But, quota logic is evaluated and saved to disk immediately.  The bug was that in the event that you were using a constructed list that is not reproducible, such as when using RANDOMIZE() without a seed, then when the quota logic was called it would overwrite the information in memory with a different constructed list.
  2. In CBC/Web, a blank row of attributes would still be displayed in the choice task if it was involved in a conditional relationship and it was the attribute marked to be displayed.  A "blank" row is a row where each level of a certain attribute across the whole task is empty or 0.

    The updated code will remove the row of the conditional relationship in the case where the conditional text is to be displayed in place of an existing attribute and it will also remove the top or bottom row if the study is configured to place the conditional text as an additional row on top or bottom.

  3. When constraining utilities in OLS for CVA/Web conjoint studies, the r-squared value could be incorrect.  This has been fixed.
  4. The utility constraints algorithm in CVA/Web had an error causing incorrect indexing of utilities, which could lead to incorrect utilities.  This could occur if the final attribute had fewer levels than another attribute in the study.
Read More

20 December 2007

The following bug fixes have been made to SSI Web:

  1. ACA/Web users who have used the advanced options to skip the Importances section or customize their Importances should read this important bug notice.  These bugs have been fixed.
  2. Bug #2 in the bug notice also exists with paper-and-pencil CVA data.  Utility estimation will be incorrect if there are decimals in the imported task responses and the analysis is done on a computer with regional settings specifying a decimal place other than '.' (period).  This has been fixed.
  3. A "file not found" error sometimes occurs when testing the survey or preparing for web upload.  This has been fixed.
Read More

04 December 2007

CVA/Web and CVA/HB are now available in the SSI Web platform.  For information on CVA/Web, including purchase information, please visit the CVA/Web product page on our website.

The following enhancements have been made to SSI Web:

  1. Pasting bulleted lists from Microsoft Word will now remove the bullet points and tabs automatically.

The following changes have been made to SSI Web:

  1. Version 6.2 changed the Quota Control behavior such that respondents who restarted a survey after their quota had filled were restarted from the beginning of the survey and all of their previous data was lost.  We have changed this behavior in this version.  Now if a respondent abandons a survey and returns to take the survey (restarts), if the respondent belongs to a quota that is now filled, the respondent will be immediately skipped to the Disqualify Question associated with that quota (the question specified on the Settings tab).

The following bug fixes have been made to SSI Web:

  1. When attempting to convert an SSI Web study to version 6.2, an "Object Required" error would be shown if the study were configured to use UserNames and/or Passwords and it did NOT have a Password question.  This has been fixed.
  2. When using large passwords sets, the size of the passwords file may grow significantly each time the Questionnaire Access Settings dialog is closed or when the "Prepare for Web Upload" operation is executed.  The error "Too many fields defined" may have been shown.  This has been fixed.
Read More

12 September 2007

The following enhancements have been made to SSI Web:

  1. Passwords
    1. Improved interface for managing passwords.
    2. Up to 125 password fields (variables) may be used (before, just two passwords could be used, named UserName and Password). You may name password variables anything you choose following SSI Web question naming conventions.
    3. Each password field can be up to 255 characters in length, using numbers, letters, or even some symbols (before, the limitation was 12 alpha-numeric characters for UserName and 8 alpha-numeric characters for Password).
    4. Up to 248 additional "merged" data fields representing information you may know about respondents prior to asking this survey may be imported and are automatically merged into each respondent's data when the respondent accesses the survey. These data fields may be used during the questionnaire, such as to show the respondent's prior data, in skip patterns, etc. Because they are part of the final data set, these merged data fields may be used in Export.
    5. You may use cookies (instead of respondent passwords) to prohibit respondents from taking the survey multiple times and also to allow respondents to restart. (Of course, if respondents have set their browser to block cookies, they may complete the survey multiple times, and they cannot restart their surveys.)
    6. A new option is available to let respondents specify their own passwords. If they terminate the survey early, they can restart where they left off by retyping their customized password.
    7. Only a single password file (STUDYNAME.pwds) is created (rather than potentially many, under the old .p001, .p002, etc. naming convention). The .pwds file is saved in a Microsoft Jet Database format (Jet databases can be modified using Microsoft Access and can be accessed from Microsoft Office applications).
    8. Previously, if importing a very large number of passwords from a text file, it could take a long time to complete the operation. The new importer is much faster.
  2. Quota Control - Quota control now allows you to set multiple independent quotas. To illustrate this idea, consider quotas involving gender (male/female) and age (young/old). In the previous version of SSI Web, the quota might be established with 4 "buckets" (Male&Young=50, Male&Old=50, Female&Young=50, Female&Old=50). This example defines the composition of 200 respondents to be collected, each bucket defined by the intersection of gender and age. With v6.2, you have another option: setting two independent quota settings and check points, one for gender and one for age. The quota rules for gender might be: (Male=100, Female=100) and the rules for age might be: (Young=100, Old=100). Notice with this approach that we aren't concerned about how many young males, etc. we interview. Each variable is considered independently--we are just interested ("on the margin") in how many male vs. female, old vs. young we interview.
  3. One-Click Respondent Access - We have changed the "One-Click" respondent access (bypass page 1) method. The syntax used for the URL to bypass page 1 has been improved, making it compatible with more systems. We have also simplified the way that variables are "passed into" the survey, having you specify the variable names in a table (as part of the new passwords dialog) rather than asking you to set up individual questions on the first page to capture these passed-in variables. See One-Click Access for more information.
  4. Admin Module - The admin module has been updated, with an improved look and easier navigation. We've also provided an advanced feature for custom logins that define which menu items are available within the admin module.

The following features have been added to SSI Web:

  1. Users can now right-click a question in the Write Questionnaire dialog, revealing a menu that lets them rename the current question (or other actions). This saves the user a few clicks.
  2. Response option labels in select-type questions are now clickable (rather than requiring just that the radio-button or check box be clicked).
  3. A new Excel file format (.xml) export option permits more than 256 columns (Excel 2007 required).
  4. Explore Study Folder icon on the toolbar automatically opens the study folder in Windows Explorer. This is a convenient time-saver.
  5. Respondent's IP Address and Browser/OS information is now exportable. Collecting this information may also be turned off if there are privacy concerns (see the new Privacy tab on the Global Settings dialog).
  6. New SSI Script Functions: Browser( ), OperatingSystem( ), UserAgent( ), IPAddress( ), EncodeForURL( ).

The following changes have been made to SSI Web:

  1. In MaxDiff Scores Analysis, the 0-100 probability rescaling procedure has been modified.  Previously, the scaling was more extreme and reflected the assumption that the respondent is choosing 1 of t items in a set, where t is the total number of items in the study.  With this latest version, we now rescale to 0-100 assuming a scale factor consistent with respondents picking 1 of c items, where c is the number of items in each choice set.

The following bug fixes have been made to SSI Web:

  1. When creating a free format variable of type checkbox, and selecting the option to force the respondent to check a certain number of checkboxes within a specified range, the dialog was requiring the survey author to make the max checks GREATER than or equal to the number of response options.  It should have required the survey author to make the max checks LESS than or equal to the number of response options.  This bug was introduced in version 6.0.0 when these fields started accepting SSI Script.  It has been fixed.  (Note that a workaround was to wrap the number in SSI Script, e.g. setting Max Checks to [%5%].)
  2. ACA Calibration Concepts may show fewer attributes than the user defined for the "Number of Attributes in Each Concept" if prohibitions are in place.  If this happens, ACA OLS and ACA HB analysis (HB only with the option turned on to calibrate utilities) would report an inconsistency.  This bug was introduced in version 6.0.0 and has now been fixed.
  3. SPSS Export bug fixes:
    1. When a Checkbox variable of a Free Format question has more than 9 items and the data are exported as "Item Numbers" instead of "Checked / Unchecked", then an error would appear during export warning that the numeric value found is too wide.  This has been fixed.
    2. The Write format for variables in SPSS data files created by SSI Web were always set to floating point (F8.2).  The Write format is used when exporting to text files from SPSS. The floating point setting could cause formatting problems.  The bug has been fixed.
    3. When exporting free format check box variables to SPSS as "Item Numbers" the missing value was being recorded for check boxes that were seen and were not checked.  This bug has been fixed and now a zero is being recorded.
Read More

24 April 2007

The following bug fixes have been made to SSI Web:

  1. Within ACA if "Randomize Attribute Order in Rating and Importance Questions" was checked and if a respondent restarted the survey within the ACARAT or ACAIMP sections then some ACARAT or ACAIMP questions may be skipped and not seen by the respondent.  This bug was introduced in version 6.0.0 and has been fixed.
  2. If a respondent finished a survey, hit the back button in the browser, and then tried to submit a page from the survey, instead of seeing "Your survey has already been submitted.  Thank you.", the browser would hang and no response from the web server would be seen.  This act would then cause problems in the Online Administrative Module.  The Online Administrative Module would hang when trying to Get Data, view the Study Summary, etc.  This bug has been fixed.
  3. The CBC designer was modified slightly.  In obscure cases, designs generated in this version may not match prior versions.  Specifically, shortcuts to the complete enumeration / balanced overlap for ADM designs were enabled (previously inadvertently disabled).
  4. In CBC if you were using any of the following:
    • An Alternative - Specific CBC Design and Shelf Facing Display
    • Customized CBC questions (using Free Format) and "Align Common Attributes on Same Row" unchecked
    Then during the interview when the software should have been displaying nothing for a level, the first level of the first attribute was being shown.  This bug has been fixed.
  5. SPSS export of multi-select (checkbox) data was not being saved correctly.  When exporting a Select or Grid question to SPSS using "Item Numbers", checkboxes that were seen but not checked were being recorded as the "missing value", when they should have been recorded as a "0".  This bug has been fixed.
  6. SPSS export of multi-select (checkbox) data was not being labeled correctly.  When exporting a Grid question to SPSS using "Item Numbers", not all of the value labels were being added to each variable, and the value labels were incorrect.  This bug has been fixed.
Read More

03 April 2007

The following bug fixes have been made to SSI Web:

  1. For CBC Alternative Specific designs using Complete Enumeration, concepts with only one attribute (constant alternatives) now are given the same chance of being included in the task as all the other possible concepts.  Previously they were included before any non-constant alternative tasks.
  2. For CBC Alternative Specific designs that used Complete Enumeration, and had non-alternative specific prohibitions, a level may not have been shown even if a level could have been shown.  Now, if a level is available, a level will always be shown.
  3. Balance was fixed in MaxDiff pair-wise designs when prohibitions are present.
  4. When saving the results of MaxDiff Counts, MaxDiff Scores, ACA OLS, or ACA/HB to a .CSV file, the resulting file would be incorrect if there were commas or quotes in the text being saved.  This bug has been fixed.
  5. Sometimes the Alt-O, Alt-C, Alt-F4 and other shortcuts caused SSI Web to stop responding (or freeze up) at times on the CBC Settings, Prohibitions, and Global Format Settings dialogs.  This bug has been fixed.
  6. When using the Export Open-Ends Only feature with "Internal Interview Number" selected as the Respondent Numbering, the values being exported for the internal interview number were not correct.  This bug has been fixed.
  7. After running SSI Web for a while and doing lots of previews, SSI Web would sometimes begin to report run-time error messages and crash. This bug is described in greater detail here. This bug has been fixed.
  8. The numeric and open-end question input boxes did not align to center or right unless there was text entered into the 'Question' text field for the question.  The bug has been fixed so that the input box now aligns correctly regardless of the existence of the question text.
  9. In Print Study "Text and Settings View" if the "Include HTML Tags" option was unchecked, and the study had foreign characters in it, then the resulting *.txt file would contain the '?' characters in place of the correct foreign characters.  This has been fixed.
  10. When printing the study to a PDF, an "Unable to render HTML" error would sometimes occur with large files.  This has been fixed.
  11. If the Progress Bar was placed in the Global Header and styles were applied, the Progress Bar would not be displayed correctly.  This has been fixed.

The following changes have been made to SSI Web:

  1. The graphical next button used "<img " in v6.0.0.  In v6.0.2 the Perl code converted this to "<input type="image" ".  Now the GUI defaults to "<input type="image" ".  When v6.0.0 and v6.0.2 studies are converted to v6.0.4, "<img " will be converted to "<input type="image" " automatically.
Read More

02 January 2007

The following feature has been added to CBC/Web:

  1. Show Levels of First Attribute no more than Once per Task.  Sometimes with attributes like brand, the researcher wants each brand to appear exactly once per choice task. If you have 4 brands and are showing 4 concepts per task, this happens naturally with Complete Enumeration or Shortcut strategies--unless you have included prohibitions in your study. When prohibitions are in place, it is possible for a brand to be repeated within the same choice task. To avoid this possibility, you can check this box, and the designer will ensure that levels of the first attribute appear just once per task.

The following bug fixes have been made to SSI Web:

  1. When trying to append CBC paper and pencil data to existing data in a study the following error would occur:

    The study data file cannot be modified because it is read-only.

    Please change this file so that it can be modified and try accumulating again.

    even though the study's data file was not read-only.  This bug has been fixed.

  2. SSI Web would crash when trying to enter CBC prohibitions with the CBC base system (CBC/Web 10) license only.  This bug has been fixed.
  3. When using CBC Constant Sum where the respondent was not required to enter an answer into each concepts' input box, the respondent would not have been included in the .CHS file export if they did not answer the first concept of all of their tasks.  This bug has been fixed.
  4. If a grid question is configured for the question direction to be "Rows", then clicking on the pencil button for any of the grid question's row list items (to edit the list item text in the SSI Web text editor) will cause SSI Web to crash.  The same will occur when editing column list items when the question direction is configured as columns.  This bug has been fixed.
Read More

08 December 2006

The following features have been added to SSI Web:

  1. Print Study.  This feature allows you to print/save a copy of your questionnaire in two formats:
    1. Graphical View: a graphical PDF or HTML format that shows what the questionnaire looks like to the respondent as displayed by a web browser, with question name labels alongside each question for easy reference.
    2. Text and Settings View: a text-only format intended for internal review that includes many details about each question: question text, response options, answer settings (such as min, max, number of responses, randomization, etc.), skip logic, lists, constructed list logic, attribute and level lists, and conjoint settings.
  2. Quick-Print Functionality. In the previous version of SSI Web, there were many dialogs that contained information that we have wanted to be able to print or save to file for further review. These areas that now are quick-print enabled are: skip logic, lists, quota control, CBC conditional text and conditional pricing, prohibitions, conjoint attributes and levels, and the question list dialog.
  3. Enhanced SPSS Export. We now can export Variable Labels (question text) using text in your Header 1, Header 2, or Question text fields (we strip the HTML). You may choose whether to export multi-select questions as either checked/not checked fields or item numbers (previously, only checked/not checked was offered). With this new version, we also export variable names longer than 8 characters to SPSS, rather than truncating them.
  4. Improved Internationalization. In the previous versions of SSI Web, you needed to change your regional settings in Windows before working on a study that involved a different character set. This is no longer necessary, and the different character sets (such as double-byte characters to display Chinese) are displayed correctly in all dialogs and reports within SSI Web.
  5. Style Selector. A new Styles tab on the Global Settings dialog lets you select different Styles (implemented via CSS) to apply to your entire survey. This dramatically changes the appearance of your survey, creating some unique and professional displays. Select from the styles currently shipping with the software, or develop and import your own styles (by following our annotated CSS file format and editing it to your specifications). We have tested our style sheets across all supported SSI Web question types (including conjoint and MaxDiff).
  6. Graphical Next Button. Allows you to specify your own graphic (rather than just text) to use for the "Next" button.
  7. New Scriptable Settings Fields. In the previous version of SSI Web, you could provide dynamic fields (scriptable with SSI Script commands) instead of fixed values for settings such as the Min and Max permitted in a numeric question, or the total points to be allocated in a constant sum question. In v6, we have expanded the scriptable fields to include number of items to rank in a ranking question, and the min and max checks permitted in a checkbox question.
  8. New SSI Script Functions. QuestionName, DisplayTotals, SortbyLabel, PageNumber, TotalPages, NumChecked, TextExists
  9. Ability to Customize Look of Online Administrative Module. This is an advanced area, but if you follow our instructions for creating a .CSS file and uploading it to your graphics folder, you can put your own company's logo and title on the administrative module rather than Sawtooth Software's.

The following features have been added to CBC/Web:

  1. Dual-Response None Format. Some researchers are including the "none" alternative as a second-stage question in CBC studies. Respondents first choose among available alternatives (a standard CBC question, without the "none"), and then specify directly below whether they would in reality buy the product they just selected. The dual-response "none" increases the statistical efficiency of the study and leads to a stronger "none" parameter (which many researchers suggest is more realistic). CBC/Web v6 supports this questionnaire format. If using Sawtooth Software tools, estimation must be done with CBC/HB v4.
  2. Constant Sum Format. For some industries (such as pharma), researchers often employ constant-sum (allocation) questions in CBC. Respondents are asked to allocate some number of points (such as 100) across available alternatives, to reflect, for example, the probability of prescribing each medication to a specific population. CBC/Web v6 supports this questionnaire format. If using Sawtooth Software tools, estimation must be done with CBC/HB.
  3. Enhanced Prohibitions (between-concept, n-way prohibitions, null levels). The dialog has been improved to more closely resemble our SMRT software product (an older generation conjoint platform) for checking prohibitions. Additionally (for those licensed with the Advanced Design Module), we now support prohibitions for levels between concepts, rather than just levels within the same concept. This allows the researcher to specify (for example) that a specific brand-price combination cannot appear in competition with another brand-price combination. For Advanced Design Module users, up to n-way prohibitions (where n is the number of attributes in your study) can be specified, for both between- and within-concept prohibitions (previously, prohibitions between only two levels could be defined). Between-concept prohibitions are less detrimental to design efficiency than the traditional within-concept prohibitions. Also, if you are including null (often as "blank") levels within your attributes, you may now prohibit too many null levels from appearing within the same concept.
  4. Enhanced Test Design (logit report based on simulated dummy data, D-Efficiency). Previous versions of CBC/Web have used a test design algorithm based on OLS principles that does not consider that concepts are grouped in sets (wherein, for example, level overlap matters). While this tool has been quick and handy, it doesn't let you test advanced designs very competently (such as those with interaction effects or alternative-specific parameters), and it doesn't support the industry-accepted standard of simulated respondent data followed by computing relative D-efficiency. The new designer automatically simulates random respondent answers and produces a logit report (allowing you to study the pattern and magnitude of standard errors for the parameters) along with a relative D-efficiency statistic.
  5. Import/Export Designs. In the past, the only way (other than through a separate utility) to view your design prior to fielding was to import dummy respondent data via the paper-and-pencil module and then export the data as a .CHO file. Also, there wasn't a way (again, other than through a separate utility) for researchers who had developed a design elsewhere to import that design for use in CBC/Web. Both capabilities will be supported in v6, using an easy, rectangular Excel-compatible file format (.csv).
  6. Short Labels for Analysis (applies to attributes and levels). This allows you to assign short labels for use in subsequent part-worth estimation and market simulations (through SMRT).

The following features have been added to ACA/Web:

  1. Option to Drop Importance Questions. In two recent research studies, we found that dropping the importance questions resulted in better share predictions for holdout choice tasks using ACA/HB estimation. If sample size is relatively large and share prediction accuracy (rather than individual-level hit rate prediction) is the goal, then our research suggests that dropping the importance questions is the better way to go. If you drop the importance questions, prior importances (for purposes of designing the conjoint pairs) can be taken from previous respondents' data, or can be set to be equal.
  2. Ability to Customize Importance Questions. Of the sections in the ACA survey, the self-explicated importance questions are perhaps the weakest link. One of our clients had a need to collect self-explicated importance information through a constant-sum question (this client was surveying sophisticated financial analysts who were adept at allocation tasks) rather than using ACA's standard importance question format. In v6, we have made it possible for you to ask importance judgments using constant sum or any other question type that saves positive numeric values.
  3. Enhanced Randomization Options. Previously, you could not randomize the order of presentation of attributes (in the importance and ratings sections), and attribute levels within the level ratings section. That functionality is now available. Also, attribute positional order in the conjoint pairs section previously was always sorted by prior attribute importance (more important attributes appeared above less important attributes). You can now turn off that sorting if you prefer that attribute position follow natural order. (For example, if brand is the first attribute in your attribute list, it will always appear as the first attribute in a conjoint pair involving brand.)
  4. New Prohibitions Dialog. We have updated the prohibitions dialog to more resemble that of our older-generation SMRT conjoint platform (ACA for Windows). This dialog was generally preferred by users.
  5. ACA/HB Integration. If you own ACA/HB v3, its capabilities are now integrated within SSI Web's menus (or you can continue to use ACA/HB as a separate program).
  6. New SSI Script Functions. ACABestLevelLabel, ACAMostImpAttLabel, ACAStrictImportance
Read More

Version 5 History

22 March 2007

The following bug fixes have been made to SSI Web:

  1. The Grid Question Numeric Variable dialog box was not saving changes to the number of decimals.  This only happened if the the checkbox indicating whether to allow decimals had not been clicked before pressing OK.  Therefore, when you turn it on it would be saved, but re-entering the dialog and modifying the number of decimals and clicking OK would not have saved the new number of decimals.  This bug has been fixed.
  2. After running SSI Web for a while and doing lots of previews, SSI Web would sometimes begin to report run-time error messages and crash.  This bug is described in greater detail here.  This bug has been fixed.
  3. If a CBC study has fixed choice tasks that have missing answers and are being imported using CBC paper-and-pencil, then a "type mismatch" error would happen.  This bug has been fixed.
  4. Select, Grid, Constant Sum, and Ranking Question dialogs do not show the correct list randomization controls when the dialog first loads up with constructed lists.  The predefined list controls were being shown.  This bug has been fixed.
  5. The OK and Cancel buttons may disappear from the MaxDiff Scores Settings dialog and MaxDiff Cho Export Settings dialog if the Windows DPI setting is not the standard setting.  This bug has been fixed.
Read More

24 April 2006

This bug has been fixed.

The following bug fixes have been made to SSI Web:

  1. If a respondent chose the "none" option for a fixed task in a CBC study then the following error would occur when exporting a CBC .CHO file:

    Unable to resolve answer for a CBC Fixed task.  Make sure that you have the correct design file, data file, and study.

    This bug has been fixed.

  2. Free Format Variable Checkbox type was enforcing the "Force respondent to check a certain number of boxes" (minimum and maximum) even though the option was not checked.  This bug has been fixed.
  3. The Grid Question Numeric Variable dialog box was not saving a change to the number of decimals.  This only happened if the checkbox indicating whether to allow decimals had not been clicked before pressing OK.  Therefore, when you initially checked allow decimals it would be saved, but re-entering the dialog and modifying the number of decimals and clicking OK would not have saved the new number of decimals.  This bug has been fixed.
  4. In CAPI mode it was possible for existing individual data records to be overwritten by new data records.  The likelihood of this happening would increase as the number of data records on a single CAPI interviewing computer would increase.  (In CAPI mode Internal Respondent Numbers are randomly generated.  If the random number conflicted with one already in use the existing data would be overwritten).
Read More

05 April 2006

The following bug fixes have been made to SSI Web:

  1. If fixed tasks were exported (using "File | Export All...") from a CBC study using random concept sorting, then the fixed task answers exported were not the correct answers when compared to the original fixed task design (in the SSI Web program).  The answer exported was the concept number that the respondent chose after the random concept sorting had been applied.  This bug did NOT affect the .cho export.  This bug has been fixed.
  2. SSI Web would overwrite other Sawtooth Software product user ids when imported through SSI Web.  No message boxes were shown to notify the user of the user id replacement or to ask if the user id should be replaced.  This has been fixed so that SSI Web will never overwrite non-SSI Web product user ids.
  3. Importing CBC paper-and-pencil data with respondent numbers greater than 1,679,615 would corrupt the respondent number.  Only the respondent number gets corrupted; all other data in the record are not affected.  The system will still not handle respondent numbers larger than 1,679,615 but now if the import encounters one it causes an error and the import is stopped.
  4. On some web servers the MaxDiff/Web design file header was not being read properly.  The result was that before the first MaxDiff set was shown the following error would be shown:

    Error:
    Exercise ID in MaxDiff design file (STUDYNAME_MXD1_1142613551.cgi) does not match exercise ID in the QST file.

    This problem has been resolved.

  5. If multi-line SSI Script (such as unverified perl) is contained in CBC conditional text, when the design is generated, the program will usually stop responding.  In some instances, it is possible that the design will generate correctly but incorrect or unexpected results may occur during respondent interviews.  This bug has been fixed.
  6. Change for MaxDiff score estimation: the way memory is managed during estimation has been updated. Large datasets (many parameters and/or many respondents) under some systems would run more slowly prior to this update.
Read More

26 January 2006

The following bug fixes have been made to SSI Web:

  1. When trying to accumulate CBC Paper-And-Pencil data an error message similar to the following was being shown (even though a valid CBC license was present):

    The number of current respondent data records (103) is greater than or equal to the number a demo study permits (5).  No data records will be accumulated.

    This error has been fixed.

  2. Multiple operators in SSI Script may evaluate incorrectly.  For example,

    (18 / 6 / 3) evaluated as (18 / (6 / 3)) instead of ((18 / 6) / 3)

    (18 / 6 * 3) evaluated as (18 / (6 * 3)) instead of ((18 / 6) * 3)

    This behavior was experienced with the division, subtraction and mod operators.  This has been fixed.

Read More

28 December 2005

MaxDiff/Web Demo is now available in the SSI Web platform.  Be sure to read the release notes for SSI Web 5.4.0 below.

The following bug fixes have been made to SSI Web:

  1. SSI Web 5.4.0 did not allow creating CBC Paper & Pencil files.  This has been fixed.
Read More

14 December 2005

MaxDiff/Web is now available in the SSI Web platform.  For information on MaxDiff/Web, including purchase information, please visit the MaxDiff/Web product page on our website.

This version uses new User Ids for all product components (CiW, ACA, and CBC).  Before you install this version, contact Sawtooth Software to receive new compatible User Ids.

Authoring surveys using Windows 98 and Windows NT are no longer supported.  SSI Web 5.4.0 will not install on a Windows 98 or a Windows NT machine.

The following enhancements have been made to SSI Web:

  1. The maximum number of concepts per task in a CBC ADM study has been changed from 101 to 251 (250 concepts if the none option is included, 251 concepts without the none option).
  2. The maximum number of levels in a CBC ADM study has been changed from 254 to 250.

The following bug fixes have been made to SSI Web:

  1. Exporting either the "Start Date/Time" or the "End Date/Time" variable to the Excel (.xls) file format on a computer with the operating system's regional settings configured to use a comma as the decimal character would create a file that could not be opened by Microsoft Excel.  When attempting to open the file in Excel, an error would display the following message:

    Problems came up in the following areas during load:

    Table

    This file cannot be opened because of errors.

    This bug has been fixed.

  2. Uploading CBC design files in ASCII text mode to a UNIX server would corrupt the design files.  A check has been added to make sure that the design files have not become corrupt by uploading in ASCII mode.  SSI continues to recommend that all files be uploaded to the server in binary mode.
  3. If the Quota Cell Qualification check point was moved (after data collection has begun) to a point further along in the questionnaire list, there was the possibility that some respondents had data erroneously removed and stored in the Limbo data folder.  This has been fixed.
  4. In the Online Administrative Module, under View Data, a data record would have JavaScript = Yes even if they did not have JavaScript enabled.  The raw data file correctly stored "sys_JavaScript, 0" but the Online Administrative Module would incorrectly interpret that and display JavaScript = Yes.  This has been fixed.  This bug was introduced in version 5.0.0 of the software.
  5. In Constant Sum and Grid Constant Sum questions, if the constant sum "min" (under the advanced tab) equals the constant sum "total" (on the main dialog), e.g. both are 100, the verification would not take place to see if it adds up to the proper range.  This has been fixed. This was introduced in version 3.0.
Read More

28 July 2005

Due to our ongoing development to better support internationalization, this version may not install correctly on Windows 98.  If you use Windows 98, please contact Sawtooth Software and we will guide you through the installation process.

The following enhancements have been made to SSI Web:

  1. The software capacity has been increased for Academic Licenses.  This includes the number of attributes in CBC and ACA (now 10), the number of CiW questions (now 500), and the number of respondents (now 250).
  2. Saving to a network drive should perform faster.
  3. Exporting should perform faster.
  4. Study Check, Prepare for Web and Preview should perform faster.

The following SSI Web bug fixes have been made:

  1. If a respondent tried to log into a survey after they had already completed it, and the data record had been accumulated, the respondent previously would see "Sawtooth Software Error #9".  This has been corrected, and the respondent now sees "No additional interviewing permitted using this username-password combination."  (This bug was introduced in SSI Web v5.0.)
  2. If an attribute level in ACA came from a constructed list, then it would be possible that the limbo data cleaner would clean data that it was not supposed to.  This has been fixed.
  3. When naming or renaming a question name, list name, or quota cell name, invalid characters were allowed if the characters were pasted using a keyboard short cut (Ctrl + V) or using the mouse right click menu.  If you try to open a study containing such a name, the study will not open.  A message is shown that a name was encountered with invalid characters.  Please contact Sawtooth Software for help opening the study.
  4. Previously, the list of export symbols saved in the .ssi file would double each time the write questionnaire dialog was closed without saving changes.  This caused the .ssi file to potentially become very large and take a long time to load and save.  This has been corrected.
  5. A compile error similar to the following:

    Error Question 'Grid1' Grid Rows - Row 9 Settings - Combo Box Response Options - The list 'Grid1' ComboList8 does not exist.

    would occur if a grid question with combo box type ranking variables had been copied from one study to another study due to the list associated with the combo box type ranking variable not being copied along with the question.  This has been fixed.

  6. We've fixed the ROUND() function in SSI Script so that trailing zeros are not improperly truncated.  For example ROUND(5.1, 2) now correctly returns 5.10 rather than 5.1.
Read More

19 April 2005

The following features have been added to SSI Web:

  1. Quota Control.  Quota Control can close the survey to new respondents after a certain number of completes.  Quota Control may also be used to achieve a sample reflecting known characteristics of a population: representative proportions of male, female, low income, high income, etc.  As an example, once enough "male" respondents have completed the survey, new males will not be permitted to fill out the survey (once gender is determined).

The following features have been added to CBC/Web:

  1. For the Advanced Design Module, we have increased level capacity per attribute from 100 to 254.  Though we think it is a rare event that would call for so many levels of an attribute, and we are cautious about the analysis implications, some clients have asked for this capacity.

The following SSI Web bug fixes have been made:

  1. Start Date / Time and End Date / Time values did not export to the .CSV or the .TAB file formats correctly.  They exported just a single digit "2" or "4".
  2. When a numeric-type answer was not answered by the respondent (either because the question was skipped by the respondent, or because the question was not required), the answer was exported as a zero (0) instead of being an empty cell when exported to the "Microsoft Excel 2002 or Later" file format.
  3. When converting studies to version 5 from prior versions, if there was a numeric-type question with decimals allowed, and the min or max had a decimal in it, the HTML created for that question would have an incorrect text box length.  For example, if the Min was 0 and the Max was 10.5 then the input box created in the HTML would only allow two numbers to be typed in.  10 or .5 could be entered but not 10.5.
  4. When Exporting to the "Microsoft Excel 2002 or Later" file format while Microsoft Excel had the file opened (or any other program), SSI Web reported that respondents were successfully added to the file.  After that, run-time error "-2147024864 (80070020)" would be displayed with the message "The process cannot access the file because it is being used by another process."  SSI Web would then crash.  No respondents would have been exported.
  5. When using some of the questions that are based on a predefined list (grid, constant sum, or ranking), if the option to randomize the order in which the list items are used in the question was chosen and then the list was then changed to a constructed list, the items were shown in the question in a randomized order when they should not have been.
  6. Using the "Randomize Response Options" checkbox on a question would give the same randomization sequence to all questions appearing at the same page position.  For instance, suppose select question Q1 has response options 1,2,3,4.  Also suppose that the response options had been randomized by checking "Randomize Response Options."  If Q1 was the first question on a page the randomization might be 2,1,4,3.  If another question, Q2, was the first question on a different page with randomized response options 1,2,3,4, the randomized order would also be 2,1,4,3.  If a third question, Q3, was on a different page with randomized response options A,B,C,D, the randomized order would be B,A,D,C.  This behavior with randomization could be found in Selects, Grids, Constant Sum, and Ranking.
    NOTE: This bug was NOT in the randomization of constructed lists with the RANDOMIZE() function.  It was caused by using the "Randomize Response Options" checkbox on the question level, which is only available when using predefined lists.
  7. When using SSI Script, a script-ending marker ( %] ) without a matching beginning marker ( [% ) caused subsequent SSI Script to not be verified or translated for use on the server.  For example, the following in the body of a Text question would not work: "%] Hello [%UserName( )%]".
  8. After changing the color of Error Message text to "Dk. Blue" in the global settings dialog, SSI Web would report the following error:  "Custom Color must be a 6-digit Hexadecimal number such as "C7F2B4"".  This has been fixed and new colors were added.
Read More

11 March 2005

The following enhancements have been made:

  1. In order to better handle double-byte character sets and to enable multilingual studies the following study files now use Unicode (UTF-8) encoding:
    • STUDYNAMEqst.cgi (server study file)
    • STUDYNAMEdat.dat (study's data file)
    • STUDYNAMEctxt.cgi (CBC conditional text file)
    • STUDYNAME.txt (Open-end export)
    • STUDYNAME.csv (Export All file type)
    • STUDYNAME.tab (Export All file type)
    • STUDYNAME.lay (Export All file type)
    • STUDYNAME.att (CBC attribute information file)
    • STUDYNAME.aca (ACA attribute information file)
  2. Microsoft® Excel® file format (for Excel XP® (version 2002) and later versions) was added to the list of available export types.

The following bug fixes have been made:

  1. With some non-Western language surveys (in particular Asian languages) if a certain character was entered in the survey text error messages it could cause JavaScript verification to break.  Also if this character was entered into ACA or CBC headers or footers the resulting text would not look correct.
  2. Only the first instance of the following ACA keywords:
    • [%ACAATTRIBUTE()%]
    • [%ACABEST()%]
    • [%ACAWORST()%]
    were replaced with the appropriate text from the attribute/level text per header.  All instances are now replaced.
  3. Ranking questions counted against the CiW license size always using the size of the list (or parent list) used in the question, even if only a sub-set of those options were required of the respondent to rank.  Now, as many questions count agains the license as are ranked.
  4. If a study created with SSI Web versions earlier than version 5.0.0 had never used an Export All operation, then when the Export All option was chosen in the converted study the Export All dialog would not work correctly.  Now in this case the .CSV file format will be chosen by default.
  5. When using Japanese Attributes or Levels sometimes the Perl Scripts would report the following error when the survey is run (either in local web or on a web server):

    Error:  Problem reading QST section:

    Bad name after _ow' at (eval 15) line 1, line 14.

  6. If the entire question name was selected when editing it on the "New Question" dialog or on the "Rename Question" dialog then the backspace key would not work for deleting the question name.
  7. Pre-skips to another question on the same page would go to that question and not show any other question before it on the same page.  So if Q1, Q2, and Q3 were all on the same page and if Q2 had a pre-skip to Q3 then Q1 would not be shown.
  8. When an attribute was deleted in CBC when a CBC partial profile design had been specified (a feature of the Advanced Design Module), multiple errors could happen, perhaps in combination:
    1. If randomizing a subset of attributes, the attributes being randomized could be changed erroneously without a warning
    2. The design could change from partial profile to traditional without a warning
    3. The design could change from full profile to traditional when it should not have changed
    4. The design could not change from full profile to traditional when it should have changed
  9. The data accumulation process that is invoked by the "Get Data" option within the Online Administrative Module has been improved.  Previously there was a possibility that if a server error occurred during the accumulation process that the data being accumulated could be lost.  The data accumulation process has now been improved to minimize any possibility of data loss in the event of a server error.
Read More

13 December 2004

The following bug fixes have been made:

  1. The following bug only affects SSI Web when running on a double-byte version of Microsoft Windows® (such as Japanese or Simplified Chinese).  SSI Web reports "WCK Error#15Nov2004" when the "Generate / Test Design" button is clicked and the study that is currently opened is in a location with double-byte characters in one of the folder names.
  2. The following bug only affects SSI Web when running on a double-byte version of Microsoft Windows (such as Japanese or Simplified Chinese).  SSI Web reports "Problem reading QST section" when you try to preview a study or run it in local web or on a web server and the study contains double-byte characters.
  3. The Hex Color codes (such as FF3012) cannot be copied or pasted.
  4. The following bug only affects SSI Web when running on some double-byte versions of Microsoft Windows (such as Simplified Chinese).  (This bug does not affect SSI Web when running on Japanese versions of Microsoft Windows.)  SSI Web reports "The offsets in the STUDYNAMEqst.cgi file are incorrect. ..." when you try to preview a study or run it in local web or on a web server.  This error will occur whether you have double-byte characters in your study or not.
  5. Version 1.5.5 studies with numeric questions would not convert to version 5.0.
Read More

15 November 2004

The following bug fixes have been made:

  1. Respondents were not able to restart their surveys if their User Name or Password was typed in upper-case (surveys actually restart, but at the beginning of the survey).  This only occured on Unix based web servers.
  2. Deleting individual data records in the On-line Data Management Module under View Data would not (if the User Name or Password was entered in upper-case) allow respondents to restart. This only occurs on Unix based web servers.
  3. Password reports could be incorrect if the User Name or Password was entered in upper-case.
  4. If the survey error messages were customized, and hard returns were entered into the error messages, then JavaScript error verification would break. This would enable respondents to submit a page without that page being verified for respondent errors.
  5. JavaScript verification for "Other Specify" breaks on older browsers (I.E. 5.0, and some Mac browsers).  This would enable respondents to submit a page without that page being verified for respondent errors.
  6. There were a few potential problems with running Mod_Perl in version 5.0.0.  There was a potential issue with the CBC design file path, and another potential problem with the Online Administrative Module.  If you are running SSI Web 5.0.0 under Mod_Perl you need to update to 5.0.1.  Call Sawtooth Software with any additional questions.
  7. The Online Administrative Module was not compatible with older versions of Perl (v. 5.004).  The effect was that the Online Administrative Module would not work at all.  A server error would display to the screen.
  8. SSI Web reported an error when trying to create a new study, open an existing study, or trying to "save as" the current study using a network location that starts with \\.
Read More

19 October 2004

The following features have been added:

  1. Constructed Lists.  One of the most important developments in v5 is the ability to create customized lists of response options for respondents, based on answers to previous questions, or other list-building logic or randomization functions.  As one example, you might first ask respondents to indicate which of many brands they consider, then ask follow-up questions regarding each respondent's considered brands.  (If you are a previous Ci3 user, it may interest you to know that CiW can now do all of the List Building examples in the Ci3 manual, adapted for the Web interviewing environment.)
  2. Predefined lists.  With v5, all response options you specify for Select-type, Ranking, Constant Sum, and grid questions become lists that may be re-used in other questions (by selecting them from a drop-down list), or can be used to build customized (constructed) lists.
  3. More powerful Skip Patterns.  In v5, skip instructions are user-specified (using typed syntax), and therefore more flexible than before.  Previously nine and/or conditions were available per skip instruction, but now there is no limit to the conditional complexity.  Skips are now question-based rather than page-based, so one indicates questions to jump from and to rather than pages (this will solve the headaches with previous versions involving checking whether skip patterns have been broken due to page inserts or deletes).  In addition to the typical skip that is executed after a question is answered, "pre-skips" are now also offered.  Pre-skips conditionally skip a question (prior to displaying it) if a certain condition is true.  Pre-skips are very useful, and can be used to conditionally show the questions on a page that apply for a respondent.
  4. Paste lists of response options from Microsoft Word or Excel.  Many users already have their questionnaires prepared in Microsoft Word.  In v5, you can paste an array of response options from Word or Excel directly into a question or pre-defined list in one step.  This offers an enormous time savings, as previously each response option had to be copied into the list of response options one list member at a time.  For ACA/Web and CBC/Web users, you can also paste lists of attributes and levels from Word as well.
  5. "Other Specify" responses.  Previous versions of SSI Web did not formally support an "other specify" response option in Select-type questions.  In v5, you indicate which option(s) on the list are "other specify" responses, and SSI Web automatically adds a text entry field next to the option.  You can ask SSI Web to verify that if the "other specify" option is checked, the respondent must provide text.  Also, SSI Web automatically verifies that if open-end text is provided, the "other specify" option has been checked.
  6. Exclusive ("None of the Above") response options.  Previous versions of SSI Web did not formally support "none of the above" (exclusive) response options.  Users had to write skip logic (or JavaScripting) to verify that respondents did not check a "none of the above" option together with any other options in the list.  You can now easily indicate which option(s) are exclusive, and SSI Web automatically verifies that respondents do not check multiple boxes including the exclusive option(s).
  7. New interface.  We've redesigned SSI Web to consolidate the main Write Questionnaire area (the separate Layout dialog is gone).  We've worked to shave extra "clicks" from standard tasks, so you can work even faster.
  8. Partial ranking in Ranking questions.  Previously, respondents were required to rank all response options in a ranking-type question.  Now, you can specify how many options are to be ranked.  This is useful if you want to ask respondents to rank their top three brands, etc.
  9. Ability to Randomize Pages.  In addition to being able to randomize questions on a page, v5 can now also randomize consecutive pages.
  10. SSI Scripting.  A variety of mathematical and string operation functions are available within "SSI Scripting."  SSI Scripting offers programming instructions that can be executed most anywhere within the software.  You can create new variables as hidden variables and modify them using mathematical functions with SSI Scripting.
  11. Cleaning of "limbo" data.  When respondents back up in a questionnaire and take a different path forward (due to skip patterns), v5 now deletes the data that are no longer valid ("limbo" data).  For example, consider a questionnaire in which a respondent branches one way or another depending on company size.  Consider a respondent that answers a few questions in the "large company" branch, and then backs up and changes the company size answer to "small company," then completes the questionnaire down the "small company" branch.  SSI Web removes any answers in the "large company" branch of questions, recognizing that those data are no longer valid.
  12. New Terminating/Link question type.  This new question type makes it easier to link multiple surveys together, as it leverages the standard Next buttons in the survey and re-directs respondents to a user-specified path, rather than relying on customized hyperlinks (such as a "Next" button graphic you have captured).
  13. Ability to customize the look of SSI Web surveys using CSS hooks and Style Sheet templates.  Cascading Style Sheets let you modify the appearance of surveys beyond the look provided by standard font and color selections.  With v5, we have provided "key word" hooks into nearly all aspects of SSI Web questionnaires (headers, footers, question text, table parameters, etc.).
  14. Ability to repeat the "column labels" in long grid-based questions.  For example, after every 10th row in a ratings table, you may wish to repeat the scale header text used in the columns.  That way, if respondents need to scroll down a large table, you can periodically repeat the column labels so that they do not have to scroll to the top of the table to review them.
  15. Improved method of launching Simple Browser.  Simple browser provides a method for removing the browser's standard buttons.  However, many respondent's browsers have interpreted the Simple Browser window as a pop-up to be blocked.  We now get around pop-up blocker software by having Simple Browser create its new child window based upon the click of a button on an introductory page rather than spawning an unsolicited pop-up like in previous versions.
  16. Faster performance on server with Mod_Perl.  SSI Web v5 is compatible with Mod_Perl.  If you have Mod_Perl installed on your server, your surveys will experience a significant performance improvement, allowing more respondents to take surveys at the same time.
  17. Easier preparation of files for server deployment.  With v5, the software automatically prepares the folder and sub-folders required for web installation for you (rather than you bringing the files together manually from multiple places).  SSI Web creates a "Web Upload" folder within your study folder that contains all files needed for web interviewing.  In most cases, this means you can simply drag-and-drop the contents of the "Web Upload" folder onto your server.
  18. Totals row/column available within grids using numeric questions.  This is a non-editable field that automatically sums up the values that a respondent enters across a row or a column in a grid.  Previously, totals rows/columns were only available within constant sum questions.
  19. Improvements to the Online Administrative Module.  These include:
    • New "Crowd Control" feature (ability to close the survey after a certain number of completed records).
    • New Test Survey Mode (including Administrative Module tools for managing test data).
    • Response option labels are now provided for select-type questions in marginals.
    • Ability to view/delete specific records.
    • Ability to download and analyze incomplete data without removing it from the server.
    • Condensed study summary table that is easier to read and understand.
    • More direct method for accumulating/downloading data.

The following bug fixes have been made to ACA/Web:

  1. Modified the way levels are selected to increase the randomness of the combinations of levels selected within attributes.  Previously if you had attributes with a large number of levels (more than 5), you could notice patterns in the levels selected across pairs for that attribute.  This did not affect the estimation of the part worths.  For an individual, the design was not as efficient as we intended.
Read More

Version 4 History

05 December 2008

  1. Bug: Importing Latent Class runs from disk could skip the first respondent. This problem was introduced with v4.18.6.
Read More

Version 3 History

13 January 2005

password: Gollum

The following bug fixes have been made:

  1. The "Specify Prohibited Pairs" dialog box may not show the correct attribute and/or level text for the prohibitions specified in the study if the attribute and/or level text contains tab characters or if it is multiple lines.
  2. SMRT may report the following error when trying to import a STUDYNAME.cho and STUDYNAME.att file exported from an SSI Web study that has level text that is more than 256 characters long and is more than just one line (these conditions will not always cause this error):

    "Error:  The file 'STUDYNAME.att' contains information in the '/@ section' that is inconsistent with the 'level information section' at the beginning of the file."

  3. If passwords were entered in upper case by the respondent then that password would not appear in the password reports in the admin module.
  4. When ACA utilities are calculated from SSI Web where any of the attribute text has more than one line then the resulting .aca file will be corrupted and will not import into SMRT and will give an error message when you try.  Now multi-line attribute text is collapsed into one line when it is exported to the .aca file.
  5. SMRT reports the following error when trying to import a STUDYNAME.cho and STUDYNAME.att file exported from an SSI Web study that has attribute text that is more than just one line:

    "Error:  The file 'STUDYNAME.att' contains information in the '/@ section' that is inconsistent with the 'level information section' at the beginning of the file."

Read More

23 February 2004

The following bug fixes have been made:

  1. Bug fixed in CBC Conditional Relationships.  Studies with a large amount of Conditional Relationship text would cause the *.cho file to be created incorrectly.
Read More

20 January 2004

The following enhancements have been made:

  1. New ACA and CBC Tutorials.
  2. Password reports in the Online Administrative Module are now more flexible.

The following bug fixes have been made:

  1. The Online Administrative Module now accumulates all data and does not use the STUDYNAMEqst.cgi file to filter out responses.  Previously if questions were removed from the STUDYNAMEqst.cgi and then the data was accumulated there was a possibility that data would be lost.
Read More

11 November 2003

The following features have been added to SSI Web:

  1. We have added new capabilities for CAPI data collection.  CAPI data collection lets you easily set up Web studies to run on stand-alone PCs or laptops for data collection (that are not connected to the web).  We hope this new functionality makes SSI Web even more useful than before, allowing users to use the software in more traditional interviewing environments.

The following features have been added to CBC/Web:

  1. For the Advanced Design Module, we have increased level capacity per attribute from 15 to 100, and have increased the number of concepts that can be displayed per task from 16 to 100.  These changes are particularly useful for packaged goods and beverage research, where it is common to have dozens of brands or SKUs.
  2. Conditional Pricing and Conditional Display capabilities.  Users can now specify conditional pricing tables which enable them to create customized price ranges for certain brands (for example), or combinations of attributes (such as brand/package size combinations).  Prices can be made to vary (using conditional "look up" tables) on the level combinations of up to three other attributes.  Conditional Display lets users display a graphic (or other HTML element) in place of (or in addition to) an attribute's regular level text.  Conditional display can vary these conditional elements based on the level combinations of up to four other attributes.
  3. Ability to suppress concept randomization.  Some researchers have wanted their CBC tasks to always show a particular brand in a certain concept position on the screen.  The default behavior in CBC is to randomize the position of levels on the screen.  We've added the capability to suppress that randomization and show the concepts in a specific order (the natural order) for a user-specified attribute, such as brand.  Thus, brand 1 will always display in the first concept position, brand 2 in the second position, etc.  Another new feature randomizes the concept position once for each respondent (for a user-specified attribute), but holds that random order constant across all tasks for that respondent.  This helps control order bias across respondents without making respondents re-orient themselves to the new positions of, say, brands on the screen within their respective surveys.
  4. Shelf-facing display presentation.  We've added a new capability to the Advanced Design Module for showing realistic shelf-facing displays, like a shopper might see in a grocery store.
  5. Paper-and-Pencil Data Collection.  Users can now use CBC/Web software for implementing paper-and-pencil CBC studies.  The menu items are nearly identical as for CBC for Windows (SMRT).  Also, we've changed the naming convention from "beginning questionnaire version" to "design seed" (though the functionality is the same).  In CBC for Windows (SMRT) the beginning questionnaire version really was a design seed, and we've clarified that terminology both in SSI Web v3.5/v5.0 and in the newest version of SMRT to reflect that.  Now, questionnaire versions always start with #1, but the designs can vary depending on the design seed.
Read More

13 January 2005

password: glacier

The following bug fixes have been made:

  1. SMRT reports the following error when trying to import a STUDYNAME.cho and STUDYNAME.att file exported from an SSI Web study that has attribute text that is more than just one line:

    "Error:  The file 'STUDYNAME.att' contains information in the '/@ section' that is inconsistent with the 'level information section' at the beginning of the file."

  2. SMRT may report the following error when trying to import a STUDYNAME.cho and STUDYNAME.att file exported from an SSI Web study that has level text that is more than 256 characters long and is more than just one line (these conditions will not always cause this error):

    "Error:  The file 'STUDYNAME.att' contains information in the '/@ section' that is inconsistent with the 'level information section' at the beginning of the file."

  3. When ACA utilities are calculated from SSI Web where any of the attribute text has more than one line then the resulting .aca file will be corrupted and will not import into SMRT and will give an error message when you try.  Now multi-line attribute text is collapsed into one line when it is exported to the .aca file.
  4. If passwords were entered in upper case by the respondent then that password would not appear in the password reports in the admin module.
  5. The "Specify Prohibited Pairs" dialog box may not show the correct attribute and/or level text for the prohibitions specified in the study if the attribute and/or level text contains tab characters or if it is multiple lines.
Read More

19 January 2004

The following enhancements have been made:

  1. New ACA and CBC Tutorials.
  2. Password reports in the Online Administrative Module are now more flexible.

The following bug fixes have been made:

  1. The Online Administrative Module now accumulates all data and does not use the STUDYNAMEqst.cgi file to filter out responses.  Previously if questions were removed from the STUDYNAMEqst.cgi and then the data was accumulated there was a possibility that data would be lost.
Read More

11 November 2003

The following bug fixes have been made:

  1. ACA Web updating bug (too much weight placed on priors)
  2. Errors in %%ANLN% display variables
Read More

26 August 2002

The following enhancements have been made:

  1. Grid, Constant Sum, and Ranking CiW questions added
Read More

Version 2 History

22 March 2002

password: WQUBIM

The following enhancements have been made:

  1. Free Format CiW question added
Read More

Version 1 History

01 August 2001

The following features have been added:

  1. CBC Web added
Read More

01 September 2001

Minor bug fixes have been made.

Read More

07 December 2000

The following features have been added:

  1. ACA Web added
  2. CiW Web including select, numeric, & open-end question types
Read More

01 September 2001

Minor bug fixes have been made.

Read More