Teacher's Corner: Conducting Specification Searches With Amos
Randall E. Schumacker · Structural Equation Modeling A Multidisciplinary Journal · 2006
Amos 5.0 (Arbuckle, 2003) permits exploratory specification searches for the best theoretical model given an initial model using the following fit function criteria: chi-square (C), chi-square – df (C – df), Akaike Information Criteria (AIC), Browne-Cudeck criterion (BCC), Bayes Information Criterion (BIC), chi-square divided by the degrees of freedom (C/df), and significance level (p). When choosing the best theoretical model from the specification search results, all or only a few of the fit function criteria may suggest the best model. Substantive theory and model validation must therefore guide any model modification process. This article describes how the exploratory specification search procedure works using multiple regression, path, factor, and structural models. Given the relative ease of conducting exploratory specification searches in Amos, researchers should begin with a substantive theoretical model to avoid misuse in empirical applications.