Solved – Use of Wilcoxon test for non-normal data akin to Two One Sided T-test

equivalencenonparametrict-testtostwilcoxon-signed-rank

I'm analysing paired data for equivalency and it's not normally distributed, i.e., the difference of the paired result is not normally distributed due to, amongst other things, outliers. If it were normally distributed I would use a Two One Sided T-test (TOST).

My questions are:

  1. Can two one-sided Wilcoxon tests with suitably defined limits ("mu" in R terminology) be used just like the Two One Sided T-test approach?
  2. If so, is it possible to establish a single p-value? Or is the larger p-value for the limit specified used?
  3. Are there any published references for doing this? (I have searched but couldn't find anything immediately available, although my search may not have been thorough enough.)

Thanks in advance!

Best Answer

The short answer is yes, you can do it, since the TOST methodology is not restricted to t-tests. The p-value is the larger of the two p-values. A quick Google search led me to a methodological article (Meier U. Nonparametric equivalence testing with respect to the median difference. Pharm Stat. 2010 Apr-Jun;9(2):142-50) describing this procedure in detail.

Related Question