On choosing a model for estimating individual differences in latent growth trajectories [Full text]

Abstract

Studies on the speed at which tasks of varying difficulties are processed frequently meet methodological obstacles, such as the difficulty to achieve reliable individual differences in parameters that describe response time growth with increasing task difficulty and the lack of stable correlations between growth parameters and external variables. In this computer simulation study, we demonstrate that instability and the systematic underestimation of correlations between growth parameters and external variables can be due to the choice of an inappropriate model to describe individual growth trajectories. Furthermore, the failure to choose a sufficiently flexible function for modeling can prevent the identification of individual differences in a shape of growth. Although we discuss our results and conclusions within the context of processing speed research, they are true for any studies that involve the modeling of latent change curves.