Coping with longitudinal data in research on developmental psychopathology

Many outcome variables in developmental psychopathology research are highly stable over time. In conventional longitudinal data analytic approaches such as multiple regression, controlling for prior levels of the outcome variable often yields little (if any) reliable variance in the dependent variable for putative predictors to explain. Three strategies for coping with this problem are described. One involves focusing on developmental periods of transition, in which the outcome of interest may be less stable. A second is to give careful consideration to the amount of time allowed to elapse between waves of data collection. The third is to consider trait-state-occasion models that partition the outcome variable into two dimensions: one entirely stable and trait-like, the other less stable and subject to occasion-specific fluctuations.