An Illustration of a Longitudinal Cross-Lagged Design for Larger Structural Equation Models

Abstract

A model of HIV behavior risk is tested, using a fully cross-lagged, longitudinal design, to illustrate the analysis of larger (i.e., 3 or more variables across 3 or more time points) structural equation models. The constructs in this integrated model were (a) Perceived Risk for AIDS, (b) Decisional Balance (relative weighing of benefits and barriers of condom use), (c) Self-Efficacy for Condom Use, (d) Lifetime Number of Sex Partners, and (d) Behavior Risk. Data were analyzed from 527 women living in a New England community who completed lifestyle surveys at 3 time points. Results of the cross-lagged analysis indicated a reciprocal, positive relation between Self-Efficacy for Condom Use and Decisional Balance. Behavior Risk for HIV was associated with higher Perceived Risk, lower Self-Efficacy for Condom Use, and lower Decisional Balance. The model showed excellent fit to the data and explained substantial variance in Time 3 measures of model constructs. Interpretations, as well as strengths and limitations of the analysis, are presented.