Dear all, sometimes evaluation of (unscaled) ABE according to FDA’s code given in the progesterone guidance does not work, i.e., the software fails to converge. This is not an issue with Phoenix (you will get the same crash in SAS), but due to the over-specified model (the test-treatment is not repeated). I would suggest to submit the error obtained in the core-output to the FDA and modify the variance structure in Phoenix: Variance Structure > Random 1 > Type [Banded No Diagonal Factor Analytic (f)] > change Number of factors from 2 to 1 This parameterization in SAS-speak would be FA0(1) instead of the FA0(2) given in the guidance. If you want to avoid convergence issues in the future opt for a fully replicated three-period design (TRT | RTR) instead. If you have any ideas why the FDA suggested a partial replicate in their guidances*, please enlighten me. Good luck. * “Draft Guidance on Progesterone” (2010) and “Draft Guidance on BE Studies With PK Endpoints for Drugs Submitted Under an ANDA” (2013)