Some Simple Methods for Estimating Intraindividual Variability in Nonlinear Random Effects Models

Three methods are proposed to improve both population and individual inference within the nonlinear random coefficient regression framework by incorporating possible heterogeneity in the intraindividual variance structure. These methods extend existing variance function estimation techniques by pooling information across individuals. The methods are appropriate when it is reasonable to assume that there exists a common intraindividual variance structure. Two common situations in which this assumption may be reasonable are in pharmacokinetic modeling and in the analysis of assay data. The proposed methods are illustrated using examples from both of these fields.