String variables: identifying matches between survey responses

[Ashwini Natraj]

Hello Statalisters,
A colleague has received survey responses and suspects that some responses have been "fed" to the respondents. Is there a neat way to identify whether responses given by any two individuals are systematically the same or similar? The problem is we don't know what strings to look for otherwise we'd use general string functions.
Otherwise, I might just ask them to use plagiarism detector softwares.
Thanks so much!

[Phil Bromiley]

How are your data structured? Are row/observations the question and columns/variable the different respondents or vice versa?

Equals works fine for strings, but note that it really expects identical strings before it says they're the same.

You'd probably need to loop over each pair of respondents and find the matching answers. But how the loop or loops work would depend on how your data are set up.

If you decide to program this, work on it a bit and then ask a more specific question.

However, given the likelihood that they don't give precisely the same response (e.g., one person puts in a double space when another puts in one space, or one person adds a word), I suspect a plagiarism detector would be better. Plagiarism detectors are designed to handle the kind of small differences you should expect even with coached responses.