AR(1) Latent Class Models for Longitudinal Count Data

TL;DR

This paper introduces a novel autoregressive latent class model for analyzing longitudinal count data, providing a closed-form solution and an efficient estimation procedure, with applications in developmental psychopathology.

Contribution

It proposes a new AR(1) latent class model for count data with a closed-form marginal distribution and a quasi-EM estimation method, improving computational efficiency and classification accuracy.

Findings

Effective in recovering latent classes in simulations

Accurate classification of subjects in real data

Provides a computationally efficient estimation approach

Abstract

In a variety of applications involving longitudinal or repeated-measurements data, it is desired to uncover natural groupings or clusters which exist among study subjects. Motivated by the need to recover longitudinal trajectories of conduct problems in the field of developmental psychopathology, we propose a method to address this goal when the data in question are counts. We assume that the subject-specific observations are generated from a first-order autoregressive process which is appropriate for counts. A key advantage of our approach is that the marginal distribution of the response can be expressed in closed form, circumventing common computational issues associated with random effects models. To further improve computational efficiency, we propose a quasi-EM procedure for estimating the model parameters where, within each EM iteration, the maximization step is approximated by solving an appropriately chosen set of estimating equations. Additionally, we introduce a novel method to express the degree to which both the underlying data and the fitted model are able to correctly assign subjects to their (unknown) latent classes. We explore the effectiveness of our procedures through simulations based on a four-class model, placing a special emphasis on posterior classification. Finally, we analyze data and recover trajectories of conduct problems in an important nationally representative sample.