Bioequivalence and Bioavailability Forum

Dear Paul!

❝ Could it be possible that the Least Squares Mean Ratio can fall outside the 90% confidence limits.

No, since the CI is calculated as ∆ (in log scale) \(\mathbf{{\color{Red} \mp}} \;t_{\alpha,df}\sqrt{\frac{2MSE}{(n_1+n_2)}}\).
\(\mathit{\Delta} = \bar{x}_T - \bar{x}_R\), where \(\bar{x}\) are the Least Squares Means (SAS-lingo; textbooks: adjusted means) of Test and Reference, \(t_{\alpha,df}\) is the tabulated value of the t-distribution (at α 0.05 and n₁+n₂–2 degrees of freedom), n_1,2 the numbers of subjects in sequences 1 and 2, and MSE the estimated residual variance from ANOVA.
Note the ∓!