A confirmatory factor model for the investigation of cognitive data showing a ceiling effect: An example

Abstract

A method for addressing the consequences of ceiling effects on model misfit in confirmatory factor analysis of cognitive data is proposed. This method focuses on the reduction of variance as a major ingredient of the ceiling effect. The model of the covariance matrix is modified in such a way that it reflects the impact of the ceiling effect on variances and covariances. The method applies to models including theory-based constraints of factor loadings for investigating cognitive data. The effectiveness of the method is demonstrated in data collected by means of a measure of working memory capacity. The application of the method in combination with a confirmatory factor model that assumes working memory capacity as the major source of performance yields the expected increase in the degree of model fit.