Linking previously collected data from a participant at Time A to Time B

I am coordinating data collection for a three year longitudinal study. We are surveying participants in six month intervals using the same survey. The problem is that participants are givig conflicting responses to questions that should not have changed or changed very much. For example: we ask a woman if she is married. At time point A it may be yes, at time B it may be no, and then at time C the answer may be N/A. We don’t know if this is an issue of errors in data collection or participants are misunderstanding the question (or just lying). If the response at Time B and C is not the same as the one given at Time A then we would like some sort of “ERROR” message or notification that the answer is not consistent. And then a space for explanation if required. Is it possible to customize the software to do this?

Dear Ms. Peterson,

our software is doing this and much more already with existing functionality. The data from previous period may be stored in HIDDEN questions, and in the validation condition you may verify whether the answer is same or different. Or if you want to request explanation, you can have to questions STATUS and STATUS_CHANGE and write an enabling condition for STATUS_CHANGE as STATUS!=hSTATUS, where hSTATUS is a hidden question, same type as STATUS. Note that the interviewers don’t see the previous value, and it may be hard to guess it, especially if this is a numeric or text question.

What you will need to do though, is to identify questions in your questionnaire where you don’t expect changes, and insert the corresponding hidden questions and explanation questions.

In some cases you expect values to grow. Things like “highest attained educational level”, or “job market experience, years” etc. In which case you may want to request explanation only if the direction changes and the value decreases.

There are many variations, including indication of differences to supervisors only, which is useful to detect cheating among the interviewers (which may be another potential explanation why the answers flip in your data).

Let me know if you need more details.

Sincerely, Sergiy Radyakin

Thank you very much! This is immensely helpful!