Bioequivalence and Bioavailability Forum

Dear Io,

replicate designs were handled in SAS usually by calling Proc MIXED, not Proc GLM.

Beside commercial software like Helmut's recommendation you can try R's add-on package nlme, function lme() or use bear which has implemented the evaluation of replicate studies (but only 2-sequence replicate crossover designs). via lme().

The estimates of the parameters of interest, namely the difference T-R and the intra-individual variabilities, are the same as in SAS, FDA code.

But the degrees of freedom for the effects tested are different and therefore also the 90% confidence intervals derived. See this thread for a discussion.

With 'enough' subjects the difference is usually not so marked. But don't know how the regulatory acceptance is.

Hope this helps.