Estimating heterogeneity in random effects models for longitudinal data.

In this paper, we are interested in estimating parameters entering nonlinear mixed effects models using a likelihood maximization approach. As the accuracy of the likelihood approximation is likely to govern the quality of the derived estimates of both the distribution of the random effects and the fixed parameters, we propose a methodological approach based on the adaptive Gauss Hermite quadrature to better approximate the likelihood function. This work presents improvements of this quadrature that render it accurate and computationally efficient in the problem of likelihood approximation with, an application to mixture models, models which allow the description of coexistence of several different homogeneous subpopulations specifying the distribution of random effects as a mixture of Gaussian distributions. These improvements are based on a new choice of the scaling matrix followed by its optimisation. An application to a phase III clinical trial of an anticoagulant molecule is proposed and estimation results are compared to those obtained with the most frequently used method in population pharmacokinetic analysis. Moreover, in order to evaluate the accuracy of the estimations, an analysis of simulated pharmacokinetic data derived from the model and the a priori values of population parameters of the previous study are presented.

[1]  D. W. Schafer,et al.  Maximum likelihood computations for regression with measurement error , 2001 .

[2]  A. Hamerle,et al.  A comparison of different methods for the estimation of regression models with correlated binary responses , 2000 .

[3]  T A Louis,et al.  Random effects models with non-parametric priors. , 1992, Statistics in medicine.

[4]  G. Verbeke,et al.  A Linear Mixed-Effects Model with Heterogeneity in the Random-Effects Population , 1996 .

[5]  Standard errors for EM estimates in generalized linear models with random effects. , 2000, Biometrics.

[6]  D. Bates,et al.  Nonlinear mixed effects models for repeated measures data. , 1990, Biometrics.

[7]  D. Rubin,et al.  Maximum likelihood from incomplete data via the EM - algorithm plus discussions on the paper , 1977 .

[8]  Marie Davidian,et al.  The Nonlinear Mixed Effects Model with a Smooth Random Effects Density , 1993 .

[9]  D. Bates,et al.  Approximations to the Log-Likelihood Function in the Nonlinear Mixed-Effects Model , 1995 .

[10]  S. Zeger,et al.  A Smooth Nonparametric Estimate of a Mixing Distribution Using Mixtures of Gaussians , 1996 .

[11]  Emmanuel Lesaffre,et al.  On the effect of the number of quadrature points in a logistic random effects model: an example , 2001 .

[12]  Murray Aitkin,et al.  A general maximum likelihood analysis of overdispersion in generalized linear models , 1996, Stat. Comput..

[13]  A. Mallet.  A maximum likelihood estimation method for random coefficient regression models , 1986 .

[14]  D. Rubin,et al.  Testing the number of components in a normal mixture , 2001 .

[15]  R. Carroll,et al.  Segmented regression with errors in predictors: semi-parametric and parametric methods. , 1997, Statistics in medicine.

[16]  Qing Liu,et al.  A note on Gauss—Hermite quadrature , 1994 .

[17]  A. Agresti,et al.  Describing heterogeneous effects in stratified ordinal contingency tables, with application to multi-center clinical trials , 2001 .

[18]  G. Tutz,et al.  Random effects in ordinal regression models , 1996 .

[19]  D. Hedeker,et al.  MIXOR: a computer program for mixed-effects ordinal regression analysis. , 1996, Computer methods and programs in biomedicine.

[20]  D. Hedeker,et al.  A random-effects ordinal regression model for multilevel analysis. , 1994, Biometrics.

[21]  C. Robert,et al.  Estimation of Finite Mixture Distributions Through Bayesian Sampling , 1994 .

[22]  G. McLachlan.  On Bootstrapping the Likelihood Ratio Test Statistic for the Number of Components in a Normal Mixture , 1987 .