The latent change model generalizes the multitrait model in the modelling of test-retest effects

Abstract : The latent change model (LCM) generalises the multitrait model (MTM) applied to psychological traits measured twice. In MTM the true gain scores are assumed to be equal for all examinees. LCM assumes a single change factor, which not only accounts for the mean structure by estimating the average test-retest effect but also to what extent individual differences arise around the average effect. Analysing empirical data gathered with three mental tests from a sample of 743 candidates to the Ecole Nationale de l'Aviation Civile, we found that introducing a change factor significantly improved the fit of the multitrait model. The change variance represented 3 to 13% of the true variance of the retest variables.