ABSTRACT
This chapter demonstrates the use of a general latent growth modeling (LGM) that accounts for the dependence of data collected in a hierarchical fashion. In this chapter, full maximum likelihood and limited information approaches to modeling balanced hierarchical and longitudinal data are demonstrated and compared. A good overview of intraclass correlation estimation is given in Koch. The chapter demonstrates how covariance structure models can be formulated for longitudinal data having nested structures, and how they can be analyzed using conventional structural equation modeling (SEM) software. One extension of the basic LGM describes individual differences within separate univariate series and forms a common factor model to describe individual differences among these basic growth curves. As demonstrated in the chapter, despite somewhat cumbersome modeling specifications, modeling four levels of the hierarchy is possible within the SEM framework. More work is needed before hierarchical extensions to levels greater than four are possible.
