## FDA loves SAS code? definitely. [R for BE/BA]

Dear D. Labes,

Sorry about the code. Here is the correct one.
proc mixed data=blabla; class seq subj prd tmt; model lnAUC=seq prd tmt/ ddfm=kenwardroger; random subj(seq); lsmeans tmt/pdiff cl alpha=0.1; estimate ’ABE for lnAUC’ tmt -1 1; run;
I've changed the parameter names to fit our discussion.

» where does the code come from?
» I cannot find it in Jones / Kenward.

Ch. 7.

» And what about the DDFM=blabla? I have learned so far (in reading tons of Web sites) that there is no such thing as different denominator degrees of freedom methods in lme()?

Nope, we cannot find equivalent one, either.

» BTW: Kenward/Jones in their Proc MIXED code always use DDFM=KR, i.e. the Kenward/Roger method. Not so astonishing I think .

Yes, you're right.

» BTW2: Let me state explicitly that these discussions are not for proving you / bear wrong. I am seriously on the way to change "The power to know" to R, but only if I can fulfill the needs in statistics for bioequivalence studies (regulators bullet proof), my everyday bread and butter work.

Wow. You can choose both, can't you? If you can't, better stay with Power to know for now.

» And currently I suspect that this is not feasible for replicate studies. Or is anybody out there to prove me that black is white?

no luck so far.

All the best,
-- Yung-jin Lee
bear v2.9.0:- created by Hsin-ya Lee & Yung-jin Lee
Kaohsiung, Taiwan https://www.pkpd168.com/bear