WLSMV Estimator in Exploratory Factor Analysis: Is It Possible?

The WLSMV estimator is the recommended approach for confirmatory factor analysis with ordinal data, but can it also be used for exploratory factor analysis? In practice, most researchersrely on the default configuration of psych::fa(), which applies Minimum Residuals estimation to Pearson correlations, because the recommended alternative, WLSMV, requires complex lavaansyntax that is inaccessible to applied users. This article introduces EFA_modern(), a function in the open-source R package PsyMetricTools that wraps the WLSMV estimation pipeline into a singlecall as simple as fa(). The user specifies only the number of factors, item names, and data frame; the function handles polychoric correlation estimation, ordinal variable declaration, obliminrotation, standardized loading extraction with optional thresholding, interfactor correlations, and comprehensive fit index reporting. To evaluate whether the WLSMV approach implemented inEFA_modern() produces superior results compared with traditional estimation methods, we conducted a Monte Carlo simulation encompassing 48,000 generated datasets across 96 experimentalconditions, analyzed with six estimation methods (288,000 fitted models). The design manipulated sample size (N = 100, 200, 500, 1000), number of response categories (K = 2, 3, 5, 7), number offactors (q = 1, 2, 3), and loading magnitude (moderate: λ ∈ [0.40, 0.60]; strong: λ ∈ [0.60, 0.80]). Results indicated that the WLSMV approach, operating on polychoric correlations, yielded theleast biased loading estimates (mean absolute bias = 0.006), the most accurate fit indices, and reliable interfactor correlation recovery. EFA_modern() produced results identical to the raw WLSMVimplementation across all conditions, validating the function. In contrast, ML and Minres on Pearson correlations exhibited systematic negative bias reaching −0.063 with dichotomous items under strong loadings, a distortion that did not diminish with larger samples. An additional analysis confirmed that this bias originates from the correlation matrix rather than the fitting function: Minresapplied to polychoric correlations matched WLSMV accuracy. We provide practical guidance, a worked example, and full simulation code. The answer to the question posed by this article is affirmative: not only is WLSMV-based EFA possible, it is methodologically superior, and EFA_modern() makes it as easy as typing a single line of R code.