Multitrait-multimethod-multioccasion modeling of personality data

Abstract In the study of the dynamic aspects of personality, longitudinal data are frequently used. Longitudinal studies allow researchers to examine interindividual differences in intraindividual changes across time, to study the consistency versus situation specificity of personality constructs, and to test for person-situation interactions. Longitudinal research in personality can be enhanced by using multimethod (MM) measurement designs. MM designs utilize two or more methods for each construct at each time point (measurement occasion) to obtain multiple perspectives and ensure construct validity. Latent variable statistical models allow researchers to separate true change and variability in personality constructs from true method effects and random measurement error. Furthermore, some longitudinal MM models allow examining convergent and discriminant validity separately for state and trait components of personality constructs. In this chapter, we review available latent variable methods for multitrait-multimethod-multioccasion (MTMM-MO) data and highlight the advantages of models that use multiple indicators (observed variables) for each method and time point. We also present an application of selected models to mother and father-reports of hyperactivity and inattention in N = 798 Spanish children.