Bioequivalence and Bioavailability Forum

Hi Mahmoud,

❝ Dear Dr

❝ What did you mean

❝ ❝ This is not a boys-only club

❝ not good words for me

Well, I meant that a top level post should be addressed the entire group of forum members – not to the “Sirs” only. Might be offensive to our female members. BTW, I’m not a Dr…

❝ I ask you about if the P-value for the random effect Subject(Sequence) in the ANOVA > .05 which means that MSinter=MSintra

A nonsignificant p-value does no imply that MS_inter=MS_inter. Not sure what you mean here.

❝ 2. if CVintra > CVinter we need to find REML estimator for MSintra

❝

❝ to find the C.I for PK

Is this a software issue? In a mixed-effects model (SAS Proc MIXED or the default BE setup in Phoenix / Win­Nonlin) sometimes rarely you may run into convergence issues (i.e., a negative esti­mate of the subject(sequence) term). Which software did you use – and which code? Consider switching to an all-fixed effect model (SAS Proc GLM without the RANDOM term; in PHX/WNL delete Random 1 sub­ject(sequence) and add it to the fixed effects). Note that with such a setup you cannot estimate CV_inter any more. However, it is not necessary for BE (which is estimated from the residual error). CV_inter is only “nice to know”.

See also these threads: #10139, #12289.