Returns and Corrections to Questions

Scenario: During the application of a questionnaire, it may be possible that the interviewer should return in the questionnaire to correct the answer to a question that could have meant a branching. As the flow of the questions may have changed, some answers must be removed or kept separately for analysis proposes.

  1. In some cases, it is necessary to keep separately the answers that were replaced for analysis purposes. Is there a way to keep those initial answers that were replaced?

There’s no easy way to preserve earlier answers to questions in the export data.

Here’s the work-around that I see. The short version: export often. The longer version: if the interview is sent to the server, then it can be exported from the server. If export occurs often (e.g., daily), then you may have earlier answers in one export archive and revised answers in another export archive. Noticing that answers have changed and matching old and new answers–these are tasks left to data analysis.

Here’s another way that might be possible if changes are made to what the paradata capture. Currently, the paradata capture all interview actions, but for question answers only preserve the current state. If paradata stored both current and previous states of question answers, then survey managers/data analysts could observe all states of answers (e.g., changing an answer from “yes” to “no”).

Correction: I believe the paradata file does indeed capture answer changes so that all answers recorded are preserved. So, for example, it would capture the following sequence of actions:

  • Answer q2 as “yes”
  • Enabled q3
  • Enabled q4
  • Answer q3
  • Answer q3 changed to “no”
  • Disable q3
  • Disable q4

For more information on the paradata file format, read here. To test this yourself, enter an answer, change the answer, sync the interview to the server, export the data, and see how this sequence of events is recorded.

Also, questions that were answered but then the different path made them irrelevant often have their input preserved as long as they have not been over-written.

On the tablet.

Once the interview hits the server, I believe that only the final answers are available. In other words, the final answers are available in the microdata, and the non-final answers are only available in the paradata.

The example I am thinking of is this, handled with enabling conditions

Question 1 - response A enables question 2 and 3 (skips 4 & 5)
Question 1 - response B enables question 4 and 5 (skips 2 & 3)

If the question was answered A the first time, then B after a rejection, I’m pretty sure the data fields in all questions 1, 2, 3, 4 and 5 would be filled and the data analysis would have to take this into account.

@sergiy can you correct me if I’m wrong?

No. The values in the disabled questions will be exported as system missing values in the disabled fields. See also here:

1 Like