Extending finite mixtures of nonlinear mixed-effects models with covariate-dependent mixing weights

Wan Lun Wang, Yu Chen Yang, Tsung I. Lin

Research output: Contribution to journalArticlepeer-review

Abstract

Finite mixtures of nonlinear mixed-effects models have emerged as a prominent tool for modeling and clustering longitudinal data following nonlinear growth patterns with heterogeneous behavior. This paper proposes an extended finite mixtures of nonlinear mixed-effects model in which the mixing proportions are related to some explanatory covariates. A logistic function is incorporated to describe the relationship between the prior classification probabilities and the covariates of interest. For parameter estimation, we develop an analytically simple expectation conditional maximization algorithm coupled with the first-order Taylor approximation to linearize the model with pseudo data. The calculation of the standard errors of estimators via a general information-based method and the empirical Bayes estimation of random effects are also discussed. The methodology is illustrated through several simulation experiments and an application to the AIDS Clinical Trials Group Protocol 315 study.

Original languageEnglish
JournalAdvances in Data Analysis and Classification
DOIs
Publication statusAccepted/In press - 2022

All Science Journal Classification (ASJC) codes

  • Statistics and Probability
  • Computer Science Applications
  • Applied Mathematics

Fingerprint

Dive into the research topics of 'Extending finite mixtures of nonlinear mixed-effects models with covariate-dependent mixing weights'. Together they form a unique fingerprint.

Cite this