I am analysing a range of clinical outcomes for a drug treatment and need to compare the values pre and post treatment - I’m using excel to do this and have found that for most of my outcomes the data is not normally distributed. Sample size is 69 patients - as such id imagine I’m best presenting my results as median and IQR but just wondering what stat test is best to run? Doing this in excel.

  • $\begingroup$ Depending on what’s the aim, a paired t-test or paired Wilcoxon test may be enough. $\endgroup$
    – utobi
    Sep 26, 2023 at 4:51

1 Answer 1


Choose your test

I typically reference one of these. Save the ol' brain power for the programs.

  • $\begingroup$ That chart fails to account for the fact that nonparametric methods and semiparametric (ordinal) regressions work great for quantitative Y. $\endgroup$ Sep 26, 2023 at 10:39
  • $\begingroup$ For the original question I'd use the rank difference test. $\endgroup$ Sep 26, 2023 at 10:40
  • $\begingroup$ Definitely, definitely agree with both of you. May I follow up with ad additional parametric/nonparametric flowchart? Or should I replace the first for the complete thought process? $\endgroup$ Sep 27, 2023 at 3:45
  • $\begingroup$ I recommend started a new chart that recognizes that an interval-scaled Y is also ordinal and that nonparametric/semiparametric methods are fully efficient. Semiparametric models can also estimate differences in means/quantiles. And models contain tests as special cases; separate tests not needed. See hbiostat.org/bbr/nonpar and fharrell.com/post/rpo $\endgroup$ Sep 27, 2023 at 11:16

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.