Factor Analysis and Variance Partitioning in Intelligence Test Research: Clarifying Misconceptions

Abstract

This article addresses conceptual and methodological shortcomings regarding conducting and interpreting intelligence test factor analytic research that appeared in the Decker, S. L., Bridges, R. M., Luedke, J. C., & Eason, M. J. (2020). Dimensional evaluation of cognitive measures: Methodological confounds and theoretical concerns. Journal of Psychoeducational Assessment. Advance online publication article.

Don’t Use a Bifactor Model Unless You Believe the True Structure Is Bifactor

“…The study by Decker et al (2020) did not claim a BF model is the same as a unidimensional model as suggested by Dombrowski et al (2020). Rather, it clarified the relationship between a unidimensional model and BF model, which has been well acknowledged by other researchers, to note the inaccurate and inconsistent conclusions in previous studies for using multidimensional BF models to support unidimensional conclusions (see Table 1 of Decker et al, 2020).…”

“…Dombrowski et al (2020) suggested the study by Decker et al equated a BF model to a unidimensional model which was a major conceptual flaw because a BF model, “…should not be construed as representing a unidimensional model of intelligence” (p. 3). Unfortunately, this concern appears to be a misinterpretation as these models were not equated and clearly distinguished in the study by Decker et al (2020).…”

“…Dombrowski et al (2020) claim “...ample justification…” (p. 3) has been given in previous studies for using a BF model and SL procedure, and specifically referenced pages 92–93 in Dombrowski, McGill, Canivez, and Peterson (2019) as providing such justification. Readers are encouraged to independently verify this claim; however, the theoretical justification from this reference is difficult to discern.…”

“…The current article provides a response to concerns raised by Dombrowski, McGill, Canivez, Watkins, & Beaujean (2020) regarding the methodological confounds identified by Decker, Bridges, Luedke, and Eason (2020) for using a bifactor (BF) model and Schmid–Leiman (SL) procedure in previous studies supporting a general factor of intelligence (i.e., “g”). While Dombrowski et al (2020) raised important theoretical and practical issues, the theoretical justification for using a BF model and SL procedure to identify cognitive dimensions remain unaddressed, as well as significant concerns for using these statistical methods as the basis for informing the use of cognitive tests in clinical applications.…”

“…Dombrowski, McGill, Canivez, Watkins, and Beaujean (2020) have raised various concerns about the research claims made by Decker, Bridges, Luedke, and Eason (2020) which questioned the methodological and theoretical conclusions made in previous studies for emphasizing the general factor (“g”) in measurement models and recommending clinical applications be limited to overall IQ scores. Specifically, evidence from a systematic review found previous studies making strong supporting conclusions, for “g” have exclusively relied on only two methods (a bifactor (BF) model or a Schmid–Leiman (SL) procedure).…”