Assume I have a dataset that ranges from 2014-2022 composed of survey responses from two different groups e.g. store 1 and store 2. I want to test whether there is a statistically significant difference between the number or percentage of those that responded 'agree' of 'strongly agree' between the two groups over time. Is there a robust way I could do this? Is there a specific test I could use or that you would recommend. I've considered a chi squared test but not exactly sure if this is appropriate or whether it would test what I am after. I have the count data and the total number of respondents for reference. Thanks all.

  • $\begingroup$ Do you want to test over all years combined? Or for a difference in trend? Or for differences in specific years? $\endgroup$
    – Peter Flom
    Commented Oct 3, 2023 at 12:34
  • $\begingroup$ Apologies I wasn't specific. I'd like to test for a difference between specific years. I.e, whether there is a statistically significant difference in the number or proportion of respondents saying they agree or disagree in group 1 vs group 2 in 2014 vs 2022. Does that answer your question? Happy to elaborate further if not. $\endgroup$
    – Daria
    Commented Oct 3, 2023 at 12:45

1 Answer 1


OK, first, it sounds like you have a Likert type scale and are dichotomizing it into "agree" and "strongly agree" vs. everything else. You can do that if you want, but it's generally not a good idea.

Second, if you want to test just two specific years, you would create a subset of your data with only those years, and then do a logistic regression with the response as the dependent variable, and group, year, and the interaction as independent variables. If you dichotomize your response, the you can use "regular" logistic. If you follow my suggestion, you would use ordinal logistic.

Third, it might be better not to subset your data and to look at all years at once. You could treat "year" as either categorical or continuous, depending on what you think is going on.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.