Misspecifications that can result in path analysis structures

Abstract

Structural equation modeling provides the behavioral and social science researcher with a sophisticated methodology to investigate causal relationships among variables within a theoretical framework. However, the researcher must guard against specification error that can result from an inadequate theoretical model. The issue of specification error is a concern in any type of causal structure, from the simplest to the most complex. In this article discussion of specification error focuses on path analysis structures, a restricted type of causal model. Such structures imply rather stringent assumptions. Violations of the assumptions are discussed as possible sources of specification error in path analysis models. Alternative models are formulated and empirically estimated to demonstrate the possible consequences of misspecified models. The discussion is extended to address the issue of specification error in LISREL-type models as well.