A Factorial Evaluation of Effects of Model Specification and Error on Parameter Estimation in a Structural Equation Model.

In an experiment manipulating artificial data in a factorial design, model misspecification and varying levels of error in measurement and in model structure are shown to have significant effects on LISREL parameter estimates in a modified peer influence model (Duncan, Haller & Portes, 1971). The extent of error present in both the measurement and the model have significant direct effects on estimated parameter values, and the two sorts of error interact as well. In specification, incorrect omission of parameters seems to have more impact on parameter error than does incorrect inclusion of parameters.