A Confirmatory Factor Analysis Approach to Test Anxiety

Abstract

This article addresses the role of test anxiety in aptitude testing. Our approach is rooted in confirmatory factor analysis (CFA). We find that the usual parameter constraints used for model identification in CFA have nontrivial implications for the effects of interest. We suggest 2 methods for dealing with this identification problem. First, we consider testable parameter constraints that identify the proposed model. Second, we consider structural relations that do not depend on model identification. In particular we derive the partial factor correlation between a test and an external variable, conditional on test anxiety, and show that this correlation (a) is not affected by the choice of model identification constraints, and (b) can be estimated using true score theory.

Keywords:

• confirmatory factor analysis
• model identification
• partial correlation
• test anxiety

ACKNOWLEDGMENTS

Peter F. Halpin is now at New York University. Paul De Boeck is now at Ohio State University.

FUNDING

This research was supported by a postdoctoral grant from the National Science and Engineering Council of Canada.