Misapplications of Simulations in Structural Equation Models: Reply to Acito and Anderson.

Although simulations can be useful for illustrative purposes or for exploring the effects of sample size and population variance on model statistics, the analysis of nonstatistical issues in structural equations should not rely on simulations. Acito and Anderson's results can be derived analytically, without simulations. Moreover, because of differences in the definition of measurement and theory, their results have limited relevance to what was suggested in our original article. In this reply we discuss these issues, question the usefulness of Acito and Anderson's approach, and further clarify the basic proposition of our original article--that reliance on covariance fit as the sole criterion for theory testing may give the analyst an incomplete basis for evaluation. [ABSTRACT FROM AUTHOR]