RE: OMEGA HAS A NONZERO BLOCK

From:"Serge Guzy" Subject:RE: [NMusers] OMEGA HAS A NONZERO BLOCK Date:Mon, 7 Oct 2002 09:30:42 -0700 I would be glad to get the raw data with all the information about the structural model, Dosage information, mixed-effect type, intraindividual variance structure and of course observed data. I would be interested to try another technique(we called the MC-PEM for Montecarlo Parametric Expectation Maximization) which so far did not have problems with ill-conditioned matrices(we indeed got sometimes high correlation that were really real or an artifact but resampling techniques were able to differentiable between reliable correlation or not). I do not know to what extent what is ill is related to "poor data factor" vs. "numerical instability or difficulty". To try the exact EM algorithm we developed would may be add some useful information. Steve, let me know if there is a way to send me the whole information about the dataset you dealt with. Serge Guzy, Ph.D. Head of Pharmacometrics Xoma