Nonlinear Latent Curve, Autoregressive Latent Trajectory, and Latent Curve Autoregressive Moving Average (LCARMA) Models as Rival Longitudinal Hypotheses: A Monte Carlo Discriminant Study

• In Event: Methodological Considerations in Modeling Latent Growth

Abstract

Voelkle (2008) found that Bollen and Curran’s (2004) Autoregressive Latent Trajectory (ALT) model fits quadratic latent curve data well but is deceiving and cautioned against its use. This study examined whether the findings of Voelkle (2008) generalize to Latent Curve Autoregressive Moving Average (LCARMA) model, and whether information gleaned from cross fitting models to respective data types could aid in distinguishing these models. Findings suggest that all models can be distinguished when correctly specified for a suitable data set so long as the underlying process is not weak. Voelkle's (2008) caution against the ALT model is limited to the situation in which the mean of the quadratic process is weak relative to the linear process with which it is associated.

Authors

• Stephen A. Sivo, University of Central Florida

• Parul Acharya, Columbus State University