Estimation in Nonlinear Mixed Effects Models：Parametric and Nonparametric Approaches

Abstract

A nonparametric approach is developed to estimate parameters in nonlinear mixed effects models. Empirical Bayes estimates of functionals of the random effects are also developed. The nonparametric approach is compared with a parametric method developed by Lindstrom and Bates (1990) in real and simulated datasets. Our numerical results show that the parametric method compares favorably with the nonparametric approach even when its assumed parametric model differs substantially from the actual mixing distribution. Applications to population pharmacokinetics are also presented.