Subject: Simulation wiht off Diagonal elements
Date: Mon, 13 Nov 2000 14:32:28 -0500


I am working with Joga Gobburu (FDA), and Nick Holford (University of Auckland) using simulation to investigate the utility of including off-diagonal elements of the covariance matrix during model building. While there are not a lot of references in the literature on this particular topic, we were able to find a few authors who seemed quite supportive of the concept of including the full covariance matrix during model building.

However, we have not been able to find references which are cautious about including the off diagonal elements covariance matrix. I was told that there is a prominent pop pk modeler in Europe who councils against using off diagonal elements during model building. I understand that this person recommend including the full covariance matrix only in the last step of model building and only if it is supported by a visual improvement in the observed vs. predicted plots. I had heard that this individual has found that including these elements during model building will result in the identification of inappropriate covariates. These findings may have been from a seminar given (PAGE?) either this year or last year.

Does anybody recognize this author or the work, and if so would you please provide the name and reference to me?

If anybody knows of published work or work in progress on this topic please let me know.

Best Regards