Heteroscedastic nonlinear regression models with random effects

Abstract

We discuss an extansion of the nonlinear random effects model from Lindstrom and Bates (1990) by adding a flexible transformation to both sides of the model (see Carroll and Ruppert (1988)) and describe a procedure for parameter estimation. This method combines pseudo maximum likelihood estimators for the transform-both-sides and weighting model and maximum likelihood (or restricted maximum likelihood) estimatiors for the linear mixed effects models. A validation of this new method is performed by analyzing a simulated set of enzyme kinetic data published by Jones (1993).