Bioequivalence and Bioavailability Forum

Hi all,

follow a recent exchange with Angus McLean, here's how I understand the normal linear model.

Let's consider the case of a 222-BE study.

I will write it a little differently than C&L do, and I do that because I understand it better this way - there i no functional difference, just a matter of notation. An explanation why this is will follow.

We write
y_ijkl=intercept+Seq_i+Per_j+Trt_k+Subj_l+error_ijkl

which means: "The observation y [can be log(AUC) or log(Cmax] is the intercept plus the sequence constant of the i'th sequence which was assigned to subject l, plus the j'th period constant corresponding to the period in which y was sampled, plus a treatment constant for the k'th treatment corresponding to the treatment plus the l'th subject constant corresponding to the subject from which y was sampled plus an error".

So in a 222-BE trial with N completers we have 2 sequence constants, 2 period constants, 2 treatment constants, and N subject constants.

In the following I will look at a dataset like this