Growth Curve Analysis Using Multilevel Regression and Structural Equation Modeling
Although the growth curve model itself is not without debate we only touch upon that discussion here. In short, there is a fair body of research comparing the growth curve model to another important model class: the autoregressive model1 (see Bast & Reitsma, 1997, 1998; Mandys, Dolan, & Molenaar, 1994; Rogosa & Willett, 1985). In summary, it seems difficult to empirically discriminate between the autoregressive model and the growth curve model. The autoregressive model may be a less restrictive model for change resulting as a consequence in the estimation of more parameters and a slightly more difficult interpretation. The growth curve model is a very elegant and parsimonious model, and this probably contributed in recent years, among other things, to its
increased application. Of importance in this chapter is that the choice between analysis techniques is not such an issue for the autoregressive model, since in practice with panel data this model is almost always estimated using structural equation modeling (SEM) software. Growth curve models, on the other hand, are commonly estimated with either multilevel regression analysis (MLR) software or SEM software.