Analyzing Panel Data in Studies of Aging

Abstract

The LISREL model, recently developed by Jöreskog and his colleagues, is of great importance to students of aging and human development. Essentially, LISREL unites factor analysis and structural equation modeling. It is extremely general, permitting a wide variety of nonrecursive structural equation models while also permitting complex models involving measurement error. Thus, LISREL speaks to the psychometric concerns of those who measure the same variable over time and to the modeling concerns of those who wish to allow for lagged effects, feedback loops, and the like. This article describes the LISREL model in nontechnical terms, points the reader to the more technical literature, and provides an extended example of a three-wave, two-variable model, with and without additional exegenous and endogenous variables. The example shows how causal effects can be estimated and how these estimates are affected by assumptions regarding measurement error.