Latent Growth Curve Analysis

Learning and development are ubiquitous. When new skills are acquired, when attitudes and interests develop, people change. Measuring change demands a longitudinal perspective, with multiple waves of data collected on representative people at sensibly spaced intervals. Multi-wave data is usually analyzed by individual growth modeling using a multilevel model for change (Singer & Willett, 2003). Recently, innovative methodologists (McArdle, 1986; Meredith & Tisak, 1990; Muthen, 1991) have shown how the multilevel model for change can be mapped onto the general covariance structure model, such as that implemented in LISREL. This has lead to an alternative approach to analyzing change known as latent growth modeling. In this chapter, we describe and link these two analogous approaches. Our presentation uses four waves of data on the reading scores of 1,740 Caucasian children from the Early Childhood Longitudinal Study (ECLS-K; NCES, 2002). Children’s reading ability was measured in the Fall and Spring of kindergarten and first grade – we assume that test administrations were six months apart, with time measured from entry into kindergarten. Thus, in our analyses, predictor t -representing time – has values 0, 0.5, 1.0 and 1.5 years. Finally, we know the child’s gender (FEMALE: boy = 0, girl = 1), which we treat as a time-invariant predictor of change.