A bifactor approach to subscore assessment.

Abstract

Bifactor confirmatory factor analysis models and statistical indices computed from them have previously been used to provide evidence for the appropriateness of utilizing a unidimensional interpretation of multidimensional data. However, the ability of bifactor indices to aid in the assessment of subscore strength has not been investigated. A simulation study was conducted to relate bifactor indices to the strength of subscores corresponding to specific factors. The bifactor indices OmegaHS and ECVSS were found to be strongly predictive of subscore strength conditional upon OmegaS. The number of factors was also found to play a minor role in this relationship. Cutoffs for assessing the appropriateness of interpreting subscores were constructed based on OmegaHS or ECVSS conditional upon OmegaS and the number of factors. For low subscore reliability (OmegaS = .60), OmegaHS = .25 or ECVSS = .45 is sufficient that the subscore has a good chance of having added value (VAR > 1.1) above and beyond the total score. For moderate reliability (OmegaS = .80), OmegaHS = .20 or ECVSS = .30 is sufficient, and the role of OmegaHS or ECVSS diminishes as OmegaS increases further. A subscore having added value does not necessitate its interpretation. Instead, when subscores are desired to be interpreted, high OmegaHS or ECVSS can be considered as evidence that such an interpretation is statistically appropriate. Finally, we illustrate the use of these cutoffs with an empirical data set. When combined with prior bifactor research, this work extends a framework of using confirmatory bifactor models for dimensionality assessment.

Reliability and construct validity of the general factor of personality disorder.

“…Following best practices in applying bifactor models to psychopathology data (Bornovalova et al, 2020; Forbes et al, 2021), we calculated reliability statistics estimating the reliability, replicability, and dimensionality of the general factor (Rodriguez et al, 2016a, 2016b), computed by BifactorIndicesCalculator in R (Dueber, 2020). The following values are presented as benchmarks for reliability analyses, in lieu of specific cut-offs (Forbes et al, 2021).…”

Evaluating the Big Five as an organizing framework for commonly used psychological trait scales.

“…R packages that were in the process of preparing this article that were not listed in the Statistical Software and Packages section of Study 1 are listed here. CFA and bifactor measurement models were conducted using lavaan (Rosseel, 2012); omega reliability was estimated with BifactorIndicesCalculator (Dueber, 2021); proportion confidence intervals were estimated with binom (Dorai-Raj, 2021); and various additional psyc analyses were conducted with the aid of psych (Revelle, 2018) and GPArotation (Bernaards & Jennrich, 2005). Packages used to create the article, tables, and figures, include: bookdown (Xie, 2021a); ggplot2 (Wickham, 2016); kableextra (Zhu, 2021); knitr (Xie, 2021b); papaja (Aust & Barth, 2018); and rmarkdown (Xie et al, 2018).…”

Psychopathy checklist: Screening version: A bifactor structure for forensic and community samples.

“…Using McDonald (1999) Omega Hierarchical estimate (estimated using semTools in R; Jorgensen et al, 2018), we evaluated the factor saturation of the latent factors. We also examined factor loading patterns, estimated the percent of variance explained, factor determinancy, and estimated H using BifactorIndicesCalculator (Dueber, 2020).…”