Chapter 1: The Discrepancy Between Measurement and Modeling in Longitudinal Data Analysis

It goes without saying (but we will say it anyway) that longitudinal research plays a critical role in education, psychology, and the health and social sciences. Also indisputable is that one of the leading approaches for analyzing longitudinal data is the multilevel growth model. In its simplest form, a multilevel growth model includes fixed effects that capture the trajectory of the typical individual and random effects that capture individual variability in stability and change over time. Substantive questions addressed by the multilevel model can concern between-person differences (e.g., do individuals high in socioeconomic status show greater achievement gains over time?) as well as within-person differences (e.g., do people elevate their levels of alcohol consumption during times of stress?). Continuing advances in multilevel modeling permit ever more nuanced assessments of change over time, including the estimation of nonlinear trajectories (Cudeck & Harring, 2007), persistent teacher and school effects (Bauer, Gottfredson, Dean, & Zucker, 2013; Palardy, 2010; Willms & Raudenbush, 1989), patterns of mediation (Bauer, Preacher, & Gil, 2006; Kenny, Korchmaros, & Bolger, 2003), unobserved heterogeneity (Muthén, 2001), and carry-over effects of time-specific shocks (Kwok, West, & Green, 2007).