ABSTRACT
This chapter reviews some modern approaches to formulation and interpretation of regression models for longitudinal data. Section 2.2 outlines notation for longitudinal data and describes basic regression approaches. In Section 2.3 we describe the generalized linear model (GLM) for univariate data, which forms the basis of many regression models for longitudinal data. Sections 2.4 and 2.5 describe different approaches to regression modeling based on whether the mean is specified directly, in terms of marginal means; or conditionally, in terms of latent variables, random effects, or response history. Many conditionally specified models are multilevel models that partition the variance-covariance structure in a natural way, leading to low dimensional parameterizations.
