Abstract
The primary objective of this article was to critically evaluate the routine use of confirmatory factor analysis (CFA) for representing an individual's performance across a battery of executive function tasks. A conceptual review and statistical reanalysis of N = 10 studies that used CFA methods of EF tasks was undertaken. Despite evidence of excellent global model fit for CFA models, studies were uniformly characterized by weak correlations among EF tasks and weak to moderate levels of maximal reliability of latent EF constructs. Vanishing tetrad tests raised the possibility that individual tasks may be better conceptualized as causal rather than as effect indicators of the latent construct of EF. The use of CFA methods with EF tasks that are weakly correlated results in latent variables that have poor to modest maximal reliability. This undermines efforts to use latent variables of EF as predictors or outcomes in empirical studies and has implications for theory development. Greater attention to and resolution of the discrepancies between conceptual definitions and statistical approaches that are used to model EF has implications for the long-term value of using EF in research and clinical settings.
Original language | English (US) |
---|---|
Pages (from-to) | 69-95 |
Number of pages | 27 |
Journal | Measurement |
Volume | 12 |
Issue number | 3 |
DOIs | |
State | Published - Jul 2014 |
Keywords
- confirmatory factor analysis
- executive function
- formative measurement
- maximal reliability
ASJC Scopus subject areas
- Statistics and Probability
- Education
- Applied Mathematics