Evaluating Models of Developmental Integration in the Laboratory Rat Using Confirmatory Factor Analysis

Abstract

This study describes a procedure for constructing and evaluating developmental models by using confirmatory factor analysis which predicts covariation among morphological measures. The principal advantage of confirmatory factor analysis lies in its ability to reject poorly fitting hypotheses in favor of those better able to reconstruct observed variance-covariance structure. Conflicting causal hypotheses of increasing complexity are applied to osteometric measures taken on a sample of one-day-old laboratory rats, then evaluated for their relative ability to reconstruct observed covariance. Models derived from competing hypotheses could not be distinguished. When models are constructed after measurements are chosen, then discrimination between alternatives may require indirect comparisons. In contrast, when hypotheses of integration vary in their predictions and the models differ significantly in fit, then confirmatory factor analysis can be used to decide between conflicting hypotheses.