Appropriateness of composites in structural equation models.

The appropriateness of using composites instead of multiple indicators in a structural model of physical health was evaluated. Liang's (1986) specification of self-reported physical health is a relatively complex multiple indicators model which may not be practical in actual application. To simplify this formulation, a two-stage strategy was used. First, reliability was estimated for each composite to fix the measurement error variance and the regression of the composite on the latent variable. Second, the model was reestimated by constraining these parameters. Regression analyses were undertaken to assess the impact of using composites instead of multiple indicators. Parameter estimates for causal linkages and residual error variances based on multiple indicators approach were closely reproduced by using composites, thus providing justification for the proposed strategy.