non-parametric equivalence /non-inferiority test for ratio variable?


New Member
Good day ,

I'm comparing subjects' success rates in two conditions A and B. All subjects run both conditions, so the data is paired. Each subject runs say 8 trials in both conditions, with 0 or 1 as output. The outcome for each subject is the ratio variable "success rate" (from 0.0; 0.125 ... to 1 with fixed ratio values). I would like to use non-parametric tests.

I would like to test equivalence / non-inferiority of treatements A and B. I'm aware that StatXact offers non-parametric tests for equivalence /non-inferiority but this seems to work only for binomial samples, i.e. when "each subject is observed for a binary response twice in succession", or "each member of a given pair is measured for the binary outcome". These StatXact tests are thus based on 2x2 tables of counts - which is not my case.

Is there an alternative, in R for example?

Many thanks for any insight!

PS. I run also Wilcoxon signed rank tests with null of equivalence, and the alternative hypothesis of difference is rejected with p=0.08, but it is not the same to accept the null of equivalence and to proove the equivalence...