Dear List, The formula expression in a glmm call is typically linear, which I believe is a requirement in order to preserve the assumption of odds proportionality. Is glmm capable of fitting a non-linear function of a covariate? In medical research a popular nonlinear function is the Emax model. For instance dose effect could be linear up to a point but would plateau at higher levels. Although a polynomial in dose may be a good compromise, but occasionally results are unsatisfactory. Use of Emax models in a glmm problem has been successfully dealt with using the package NONMEM and recently WinBugs and SAS. I am curious if the non-linearity poses a theoretical challenge, a computational one or both? Thanks Farkad Ezzet Ph. D. Pharsight Corporation [[alternative HTML version deleted]]