Abstract

SILVA, Marli Appel da; ARQIMON, Irani Iracema de Lima  and  WENDTD, Guilherme Welter. Correction methods for Structural Equation Modeling statistical tests. Aval. psicol. [online]. 2024, vol.23, n.1, pp.109-120.  Epub Dec 02, 2024. ISSN 1677-0471.  https://doi.org/10.15689/ap.2024.2301.18031.11.

This study aimed to compare different statistical tests by increasing the sample size in samples that violate the assumptions of normality in Structural Equation Modeling (SEM). The following methods were compared: null hypothesis test (H0SB) — Satorra-Bentler scaled chi-square test (χ2SB); the strategy suggested by Moshagen and Erdfelder (TM&E); and the equivalence test proposed by Yuan et al. (TY). This was a simulated, exploratory study. The H0SB, TM&E and TY tests revealed distinct behaviors, suggesting that the H0SB can support smaller samples (100 cases) and reject larger samples (≥500 cases), and that the TM&E and the TY tests can support models with more cases (≥500 cases). The TY test is recommended because it presents more resources for evaluating models in SEM, although the TM&E test deserves further studies on its performance, as it could be an important additional tool.

Keywords : methods; statistical analysis; Structural Equation Modeling.

        · abstract in Portuguese | Spanish     · text in Portuguese     · Portuguese ( pdf )