Finding Missing Data in Limbo

Last Updated: 10 Sep 2014Hits: 4956
I'm using SSI Web 7 to field a survey and I have a respondent who is marked as complete but when I examine their data record, it shows that they skipped a required question. Where is the missing data?

Occasionally you might discover that a data record is missing the answer to a question.  It's always best to step through the survey using the respondents answers in test mode to make sure the question was indeed shown, but sometimes a response is "limbo cleaned" by accident.  Limbo cleaning can occur when SSI Web tries to clean up responses that are no longer valid.  This article focuses on surveys prior to version 8, where data was stored in individual files for respondents.  Version 8 changed to store data in a database, so the instructions for fixing this problem below are not applicable.  

For example, suppose you have a survey with the following question:

Q1. Do you attend college?  [Y/N]
Q2. Which college do you attend? [Enter Name]
Q3. How much do you expect to make in 5 years? [Enter Value]

If a respondent answered "Yes" that they attend college, they would input an answer for Q2, and the following is recorded in the data file:

  1. Q1,1
  2. Q2,College_Name
  3. Q3,50000

Let's assume there is skip logic in place that Q2 is only ever seen if Q1 is Yes. If the respondent backs up and changes their answer in Q1 to No, they will skip Q2 and move on to Q3.  After making the change, the data file would look like this:

  1. Q1,2
  2. Q3,50000

Removing Q2 from the data file is called "limbo cleaning" because we still store the answer to Q2 in a limbo data file.  On your server, inside the data folder, there will be a studyname_limbo folder.  Each file inside that folder will match a respondent's data file and contain any limbo-cleaned data.

Occasionally data is accidentaly removed that shouldn't have been limbo cleaned and placed into the limbo file for that respondent.  You can grab their answer and place it back into the respondent's data record to fix the problem.  The raw data files can be opened in a text editor, such as Notepad.  The one catch, however, is that you might have to change the first row of a respondent's data, which acts as a header row.  An example of the header row would look like this:

  1. 4 349 51409804917 25 606207 1409804814 QEnd

The two variables we'll focus on is the second, 349, and the fourth, 25.  The second is the respondent number, and the fourth indicates how many lines follow the header in the respondent's data record. If you add a new line to the record to replace the limbo-cleaned data, we would want to change the 25 to 26 to accurately reflect 26 lines of answers following the header.