Version History

This release includes support for the Academic Research licenses.

This release resolves installation issues with SSI Web v7.

1. Feature: SMRT has been updated to coincide with Latent Class v4.1.

1. Feature: SMRT has been updated to coincide with SSI Web 8.

1. Bug: SMaRTQue would not run from the Run command. This bug is specific to v4.20.0.
2. Bug: Field disk creation would fail. This bug is specific to v4.20.0.

1. Feature: SMRT has been updated to coincide with SSI Web 7.
2. Change: This version no longer supports Windows 2000.
3. Bug: Exporting utilities to CSV format encountered an error when a variable was coded linear.

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

1. Bug: Importing and saving Latent Class runs would not save beyond the first solution in the file. This problem was introduced with v4.18.6.

1. Change: Importing .hbu and .utl files now allows one line per respondent.

1. Change: CBC .cho export now allows merged variables to be exported as floating point numbers. Previously they were rounded to integers for compatibility with CBC v1.x.
2. Change: SMaRTQue no longer gives a warning if it cannot locate one of it's 'native' screen resolutions. The introduction of widescreens has created many new resolutions not available when SMaRTQue was written.
3. Bug: Using constraints in Latent Class on an interaction when more than one interaction was present could result in incorrect utilities or a memory error message.

1. Bug: CVA/Web licenses weren't recognized for File|New.
2. Bug: Simulator would crash if all external effects were zero.

1. Feature: CVA has been updated to coincide with SSI Web 6.4.0.
2. Change: CVA and ACA utility imports may also use a .att file instead of a .cva or .aca file.

1. Change: In conjuction with new license policy, SMRT will now activate authoring based on an SSI Web v6.x license.
2. Change: Latent Class v3 will activate based on a Latent Class v4 license.
3. Bug: When displaying standard errors in Latent Class v3, the errors for the last group were displayed for every group.

1. Feature: Randomized First Choice has new options for using independent error for attributes.
2. Change: The Help Files have been converted to a new format for use with Windows Vista.

1. Feature: When merging variables, custom segments can be created for the variables. A segment will be made for each unique value of that variable.
2. Feature: Paper & Pencil has a new export for exporting designs only.
3. Feature: Paper & Pencil has a new option to include blank lines for levels not shown.
4. Change: The .cho importer will now allow files formatted with dual-response none. They are interpreted as if no 'none' was present.
5. Fix: Double-clicking a .smt file would not load the correct study.
6. Fix: ACA attribute help text could not be removed.
7. Fix: Problem recognizing CCS license with Simulation Only studies.
8. Fix: Problem getting proper ADM size of SSI Web v3.5 and v5.0 licenses.
9. Fix: Logit could crash if missing a main effect.
10. Fix: Latent Class could crash if missing a main effect.
11. Fix: Latent Class could crash if level text exceeded 255 characters.
12. Fix: Paper and Pencil would crash if the questionnaire contains the ‘For analysis only’ mock-question creating when importing with an analysis only license.

1. Change: Management of User Ids has changed to accomodate future releases of all products including SSI Web, Ci3, CBC/HB, and CPM. Users of Ci3 will need to update to v2.6.12 to be compatible with this version.

1. Change: The software capacity has been increased for Academic Licenses. This includes the number of attributes in CBC, ACA, and CVA (now 10), the number of products in the simulator (now 100), and the number of respondents (now 250).

1. Change: CBC/Web ADM now allows analysis of 254 levels for an attribute. This is to support SSI Web 5.2.0.
2. Bug: Importing a cho file into a CBC study where respondents had different numbers of tasks would result in an invalid .dat file.
3. Bug: View/Edit data would allow users to give answers to choice tasks whose design was never generated. This would cause estimation to crash.

1. Change: Changes to .att and .aca imports to read UTF-8 encoded files. This is to support SSI Web 5.0.4.

1. Feature: A new User ID import option has been added which reads a User ID from a file.

1. Change: ASM now allows (with a warning) price attributes to be included in the Attribute-based Price/Cost tables. While this is permitted, it is generally not appropriate.
2. Bug: When importing merged variables, utility runs, or accumulating paper and pencil data, if the text file had UNIX-style formatting (single line feeds instead of carriage return/line feed pairs), the import would product incorrect results.
3. Bug: In CBC data sets with large numbers of tasks and concepts, Latent Class would be unable to produce a valid solution. A constant used to avoid divide by zero problems while estimating group membership was too large.
4. Bug: In Latent Class output, the summary of replications occasionally would not have enough width in the text, causing the numbers to run together.

1. Bug: Using Sensitivity analysis in standard mode and using Price Associations, price increments from the associations would accumulate between simulations rather than being reset. This error was introduced in SMRT 4.7.4.
2. Bug: Also using Sensitivity in standard mode with Price Associations, if the simulation method does not support extrapolation (such as Randomized First Choice), the scenario might error with a message "An unknown file tried to seek beyond its end." This error was introduced in SMRT 4.7.4.

1. Feature: Sensitivity analysis now has a new section at the bottom of the report when using standard mode. It contains a summary table for the sensitivity products x sensitivity attributes share results at the very bottom of the simulation output. This allows users to, for example, produce table of shares needed for demand curves.
2. Bug: Some errors occurring during an utility import were not being displayed. The import would fail, but there was no indication as to the problem. This error was introduced in SMRT 4.7.0.
3. Bug: When specifying cost/revenue/profit information in the ASM, the "Ignore price tables for static products" option was not being saved correctly. This error was introduced in SMRT 4.5.3.

1. Bug: Logit analysis would crash if hard returns were used in the level text. The same bug in Latent Class would cause all solutions to be reported as ill-conditioned. This error was introduced in SMRT 4.7.0.

1. Change: The simulator has been modified so that ACA, CBC, CVA and CCS (except lab versions) support up to 100 products.

1. Bug: When running Latent Class under Windows 95/98/Me, if the text output grew greater than 32KB, it would be truncated and the output would not display in the window, and not be written to the log file. Although it did not affect the quality of the estimated part worths, users could not view the log to view the details of the solution. This error was introduced in SMRT 4.7.0.
2. Bug: Occasionally while running Latent Class, a constraint on a linear attribute would result in the message "Too many constraints". This error was introduced in SMRT 4.7.0.

1. Product: Incorporates Latent Class v3.0
2. Feature: When merging variables from text-delimited files, labels may be specified on the first row of the file instead of forcing the user to enter them manually.
3. Feature: In CBC, answers to fixed choice tasks are now automatically available for segmentation, similar to the results of single select questions.
4. Change: ACA has been modified so that a 10-attribute license may have up to 30 products in the simulator.
5. Change: The attrib.tmp, logit.tmp, and *.utc files from the CBC Logit output have been removed.
6. Bug: During simulation, if all respondents were removed from a run via the respondent filter, Randomized First Choice would report a divide by zero error. This error was introduced in SMRT 4.5.
7. Bug: When interviewing using SMaRTQue, interview times were accumulated between respondents, rather than being reset for each respondent. This would only occur if the interview was set to recycle after a respondent finishes. This affects all versions of SMRT.
8. Bug: Merged variables would save as the wrong type (an imported text-delimited file would show as a SMRT .dat file, etc.). Although it did not affect analysis of the data, if the user chose to update the merged variables using the option in SMRT, the system would crash. This error was introduced in SMRT 4.6.0.

1. Feature: SMRT now recognizes SSI/Web Version 3.5 licenses.

1. Feature: Assigned Level Values for CBC Conditional Pricing is now supported in the simulator. This feature allows users to directly specify the actual price values for simulated products from the conditional pricing table rather than be limited to using level codes only for price.
2. Change: ASM has been updated to support attributes with up to 100 levels.
3. Change: The speed of simulations has been significantly improved, due to optimized code, mostly in the area of memory management.
4. Change: Because of the increase in speed, we’ve increased the number of default iterations in Randomized First Choice from 50,000 to 100,000 (to obtain slightly better precision). The user may still raise or lower this value. Even if using the same number of iterations as before, you will notice slightly different share results using RFC in this version, but that difference is due to random error.
5. Change: The speed for CBC and CVA Paper-and-Pencil accumulate data steps has been increased.
6. Bug: With the Advanced Simulation Module (ASM), specifying assigned level values from largest to smallest (for example assigning 30, 20, 10 for levels 1, 2, 3) would result in the <= and >= operators being reversed. For example, if an attribute's level values were specified as 30, 20, 10 and the product search was ">= 20", the search would be on values 20 and 10. A further example is if a prohibition was specified "<= 20" the values 30 and 20 would be prohibited.
7. Bug: In sensitivity analysis, specifying an attribute as NA would result in the product being treated as extrapolated, and therefore there would be no results. The message "Warning: there were (n) attempts to simulate on prices using extrapolated values" would be shown in the simulation results. This did not occur with ASM Exhaustive mode sensitivity. This bug is only present in SMRT 4.5.4.

1. Bug: If the study includes linear attributes, when using standard simulations, a linear attribute's level value would be incorrectly encoded. This would result in incorrect simulation results. This does not affect sensitivity analysis or product search. This bug is only present in SMRT 4.5.3.
2. Bug: If using sensitivity analysis and price associations, and the simulation method used was Randomized First Choice, the simulator would perform extrapolation checks on the price association's values, which are actually delta values which are not checked. Also, there was no extrapolation check made after price association deltas were applied, and RFC would report invalid results (extrapolation cannot be used with RFC).

1. Feature: ASM now allows attribute-based pricing. This permits users to specify that certain attribute levels are associated with different prices. These prices are added to the price attribute to determine the overall price for products. Attribute-based pricing may be used for utility, purchase likelihood, share, revenue, and profit calculations.
2. Bug: In ASM product searches, if the study includes linear attributes, and a seed product is used, the linear attributes will be seeded incorrectly. There is a high chance that the search will report that the seed is invalid. If no invalid seed message is shown, it does not mean the linear attribute seeded correctly.
3. Bug: In ASM product searches, if the study includes linear attributes, and Randomized First Choice is used, the simulator may report an extrapolation error on a valid value.
4. Bug: In ASM product searches, if the study includes linear attributes, and interaction terms are used, the simulator may report an extrapolation error on a valid value.
5. Bug: In ASM product searches, if a linear attribute’s assigned level values were changed after setting up the scenario, it could allow extrapolated values into the simulation.
6. Bug: In ASM product searches, when using the Exhaustive search algorithm, if a step value is specified, but no interpolation range, and the product specification includes an endpoint which does not fall on the step value, the search will enter an infinite loop.

1. Feature: ASM licenses have now been extended such that if ASM is the only installed license, it behaves as a Client Conjoint Simulator, being able to open conjoint studies and simulate for their runs. Also, a new "Simulation Only" study type is now available when creating new studies. This study type allows importing .hbu-formatted text files, and analyzing the utilities inside the Market Simulator. The Client Conjoint Simulator can open "Simulation Only" studies created by ASM users, but is limited to standard simulation functionality.
2. Feature: SSI Web users no longer require a SMRT Analysis User ID to analyze data in SMRT.
3. Change: The results display in the ASM has changed so that in order to show Profit, a price attribute must be indicated.
4. Change: SMaRTQue has been changed such that when running ACA studies, the likelihood of receiving a "Too many constraints" message is now reduced. This occurred when attributes adjacent to another in the attribute list had too many prohibitions.
5. Change: The .hbu export from the Utility Manager has been improved. With earlier versions, attribute and level labels were only exported as "A#L#". Now the actual labels are exported (only the first 255 characters are exported).
6. Change: The dialogs for Tables, Counts, and the Market Simulator now support resizing. They may also be maximized to fill the screen.
7. Change: Support for draw files (.dra) has been removed. Sawtooth Software has concluded that draw files have no distinct advantage over conventional point estimate HB utilities. Importing .dra files is no longer available, and simulations using draws are no longer allowed. Utility runs using draws may still be exported for use in other environments.
8. Bug: When using Sensitivity analysis in the simulator, and also using Randomized First Choice as the simulation method, an autocalibrated value was being displayed in the settings for the attribute variability multiplier. The value corresponded to only the first simulation performed in the search.

1. Product: Incorporates the Advanced Simulation Module (ASM) v1.0.
2. Change: A new autocalibrated attribute error function is being used, along with a new random number generator. Both of these will cause small differences in Randomized First Choice results compared to earlier versions.
3. Change: The RFC (Randomized First Choice) simulation method has an option for "Total Sampling Iterations" (default 50,000). This number is divided by the cases (respondents) in the utility run and the result is the number iterations needed to reach the total. For instance, with 500 cases, 100 iterations are needed to reach 50,000 total sampling iterations (50000 / 500). In the prior version, applying a respondent filter would *not* change the above calculation. This meant the Total Sampling Iterations would not be truly realized using a respondent filter. For instance, if 100 cases were excluded through a filter, only 40,000 total sampling iterations would be performed (400 * 100). The Total Sampling Iterations reported in the output was 50,000 (or the default override number specified by the user). In this version of SMRT, applying a filter will allow the Total Sampling Iterations to be reached. The Total Sampling Iterations number is divided by the number of cases that passed the filter. This means the number of iterations specified is always performed, regardless of the filter.
4. Bug: SMaRTQue calculated ACA final utilities incorrectly when using a calibration scale other than the default 0-100 scale. If a narrow scale such as 0-5 is used, final utilities will be very similar across respondents. ACA/HB is unaffected by this problem.
5. Bug: In ACA OLS utility estimation, the way utilities are calculated for attributes not taken into pairs has been improved. The new method ensures that the final importance of attributes not taken into pairs relative to those included in pairs is more logically consistent, more closely matching the approach used in ACA/HB for constraining importances to be consistent with the prior ratings.

1. Bug: When using the Generic Cross Tab export in ACA studies, if the study has an a-priori attribute and an unacceptables question is asked for that attribute, no blank field was being inserted for a level marked as unacceptable by the respondent, and so the remaining data fields for that respondent would be shifted one field to the left.
2. Bug: When using the Generic Cross Tab or SPSS export in ACA studies, occasionally preference data for a-priori attributes will appear as blanks.

1. Change: Sensitivity analysis was not taking into account that some attributes may not be included in the effects coding of the run, or that the user may have specified "NA" for the sensitivity attribute. Therefore that attribute would be iterated upon. The results for non-coded attributes would be the same for all levels of that attribute. The results for NA attributes were as if they had not been marked NA. Attributes that satisfy either condition are now skipped in the sensitivity analysis.
2. Bug: When centering products for linear terms, "NA" attributes were being centered, essentially making them no longer "NA". The closest estimation to what was happening is that they would be centered as if they were approximately a 0.
3. Bug: Randomized First Choice, when computing the differences between products, was not taking into account that some attributes may not be included in the effects coding of the run. Therefore, the value the user had entered in the product specification was still being counted in the computation.
4. Bug: Randomized First Choice, when automatically computing the attribute variability multiplier, was not taking into account that some attributes may not be included in the effects coding of the run. Therefore, the attribute error was calibrated as if the run included the attribute.
5. Bug: If using Share of Preference, Share of Preference with Correction, or Purchase Likelihood in the simulator, it is possible that extreme utility values or extreme external effects or extreme none weights would cause overflow or underflow in the processor. In the case of underflow, shares would not add up to 100 percent. In the case of overflow, shares may be output as a -1. Clicking on the cell would display cell contents of -1.#IND0.
6. Bug: If using Share of Preference with Correction, when computing the number of differing attributes to correct on, SMRT was counting a column of NA's only if the study included a 'none' option (either specified in CBC or in the imported utility run).

1. Product: Incorporates CVA/HB v1.0, hierarchical Bayes estimation for CVA.
2. Change: In CBC, when you export a .CHO file, an accompanying .ATT (attribute labels) file is now automatically generated. The .ATT file is necessary for Latent Class and HB estimation programs.
3. Change: In CBC, the Test Design function is now available for the Advanced Design Module when using Alternative-Specific plans.
4. Bug: If a SMRT 2.7 or later user selects the "Apply to all" checkbox in the Advanced Scenario Settings dialog box and applies the advanced scenario settings to scenarios created in a version prior to 2.7, the StudyName.UCS file will become corrupt. The file will lose data, contain invalid data, or become unreadable. The affected versions of SMRT are 2.7 to 4.0.0. Any SMRT component that uses the simulator may be affected. This includes ACA, CBC, CVA, and CCS (Client Conjoint Simulator).
5. Bug: CBC Paper and Pencil keeps track of the number of times each level is shown within the paper-and-pencil design, across all questionnaire versions. If that number exceeded 32,768 for any one level, then the internal counter incorrectly wrapped back to zero. To strive for balance, CBC then kept generating concepts with that one attribute level whose counter had been set back to zero, believing it had not occurred enough times in the design. This essentially turned the remaining choice tasks into "partial profile tasks" as for that affected attribute, the levels were held constant for all remaining choice tasks. We think that it is quite rare that a user would encounter this bug, but it is potentially data threatening.

1. Product: Incorporates CVA v3.0 (CVA on the Windows platform).

1. Product: Incorporates ACA v5.1 (Enhancements to ACA v5.0, such as a 'Single Select' question type).

1. Product: Incorporates ACA v5.0 (ACA on the Windows platform).

1. Change: This version supplies various enhancements to the SMaRTQue interviewing module, including the ability to have graphics in the "top/bottom/left/right" areas of the question, support for two file sizes for "zoomable" graphics, and the ability to run SMaRTQue from CD-ROM.

1. Product: Incorporates CBC v2.5.
2. Product: Incorporates CBC Advanced Design Module v2.5
3. Product: Incorporates ACA v4.5 and CVA v2.5, which allows the simulator to work with ACA and CVA data.
4. Product: Incorporates CCS v2.5.
5. Feature: The simulator now imports Individual Choice Estimation (ICE), Hierarchical Bayes (HB), Latent Class, and generic conjoint utility files.

Initial release including CBC v2.0 (CBC on the Windows platform) and Client Conjoint Simulator (CCS) v2.0