Confirmatory Factor Analysis in Kinesiology Journals with Explicit Measurement Focus

ABSTRACT

Confirmatory factor analysis (CFA) is a commonly used statistical technique. Recommendations for evaluating CFA highlight scholars should outline the expected model, conduct data screening, report model estimation and evaluation, and report key information about results to provide evidence for latent variables. The purpose of the current study was to review and evaluate the use of CFA in research published from 2013 to 2022 in the journal of Measurement in Physical Education and Exercise Science (MPEES), the most measurement focused journal in kinesiology, in our view. Results were cross-checked by examining research published in Research Quarterly for Exercise and Sport during the same time period. Strengths included providing information about the expected model, evaluating the final model, and reporting parameter estimates. Areas for improvement included conducting data screening and evaluating factor quality. Finally, recommendations for scholars to improve reporting of CFA in kinesiology are provided.

KEYWORDS:

• Factor analysis
• reporting practices
• structural equation modeling

Disclosure statement

No potential conflict of interest was reported by the author(s).

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/1091367X.2023.2270466

Notes

¹ Variance explained is calculated by squaring the standardized factor loading. However, this is only possible if the observed item is an indicator for one factor or orthogonal factors.

² Kyrgiridis et al. (2014) used principal component analysis with oblimin rotation.

³ Maïano et al. (2022) reported using algorithms in Mplus to handle missing data when using the WLSMV estimator. We assume the authors are referring to pairwise present, the default in Mplus when using WLSMV. Lyyra et al. (2015) reported using the missing data method in Mplus to handle missing data when using the MLR estimator. We assume the authors are referring to full information maximum likelihood, the default in Mplus when using ML.

⁴ Starnes et al. (2019) report using “full information maximum likelihood” in Mplus to analyze the CFA models (p. 140). We classified this as ML.

⁵ Research Quarterly for Exercise and Sport publishes research in biomechanics and motor behavior, exercise physiology, measurement and evaluation, physical activity and health behavior, psychology and social sciences, and pedagogy. However, these subsections are not noted on the online volumes nor the hard-copy volumes. Thus, prevalence of CFA by subsection could not be calculated.

⁶ We discuss Mplus because it was the most used software among the reviewed studies.

Additional information

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.