Using Multivariate Data to Structure Developmental Change
J. J. McArdle John R. Nesselroade University of Virginia
Life-span developmental psychologists have been sensitive to several key methodological issues in the domains of measurement, research design, and modeling. Among their prominent concerns are the development and elaboration of multivariate approaches to identifying and studying change phenomena. These concerns bear on the operational expression and definition of latent variables, the separation of qualitative and quantitative change, the modeling of individual differences and similarities in patterns of intra-individual change, and the assessment of treatment effects.
Data collection and analysis strategies that permit the structuring of interrelationships among many variables simultaneously have proven to be indispensable aids to understanding the nature of relevant phenomena. During the past three decades or so, methodological innovations and improvements have markedly increased our ability to structure relational matrices that contain information about development and change. Some of these methods can provide both general and individual differences information simultaneously.
In this presentation we describe some promising innovations from the growing literature on Linear Structural Equation Modeling, a group of techniques that rely on LISREL and similar computer programs (see, Fraser & McDonald, 1988; Goldberger, 1973; Joreskog & Sorbom, 1979; McDonald , 1985 Nesselroade & McArdle, 1986). The models and analyses we highlight bear directly on fundamental concepts of change. Developmental