This paper discusses two methodological issues regarding the analysis of longitudinal data using structural equation modeling that emerged during the reconsideration of the analysis of a recent study on the relationship between academic motivation and language achievement in elementary education [Stoel R.D., Peetsma, T.T.D. and Roeleveld, J. (2003). Relations between the development of school investment, self-confidence and language achievement in elementary education: a multivariate latent growth curve approach. Learning and individual differences, 13, 313–333]. The issues are related to the factorial structure of the repeatedly measured variables, and to the explanation of interindividual difference by means of covariates [see Stoel, R.D., Van den Wittenboer, G. and Hox, J.J. (2004a). Including time-invariant covariates in the latent growth curve model. Structural Equation Modeling, 11, 155–167, Stoel, R.D., Van den Wittenboer, G. and Hox, J.J. (2004b). Methodological issues in the application of the latent growth curve model. In K. van Montfort, H. Oud, and A. Satorra (Eds.). Recent developments on structural equation modeling: Theory and applications. (pp. 241–262). Amsterdam: Kluwer Academic Press. It is illustrated that standard modeling practices may sometimes lead to incorrect conclusions regarding the concepts under investigation, and that ideally alternative modeling possibilities should be considered in order to check the adequacy of the standard practice.
Read full abstract