On the use and abuse of wording factors: A systematic review and estimation of derived indices using the RSES
Loading...
Date
2023
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Several studies have proposed the bifactor structure as the optimal model to represent wording effects observed with balanced scales. This specification allows to separate variance accounted for by the attribute from the variance attributable to the measurement method, which is used to examine the substantive or artifactual interpretation of the wording effect. However, the suitability of these method factors for this use is uncertain due to the risks of overfitting and lack of stability and interpretability of the estimated parameters, aspects barely examined in previous research. In this paper, using different indices of model assessment, we examine the degree to which the attribute and wording factors are psychometrically well-defined to serve in the SEM context as predictors or criteria, illustrating their use with the RSES scale. We carried out a systematic review in four databases (Web of Science, Scopus, PsycArticles, and Google Scholar), which resulted in 71 records that reported the factor loadings of the bifactor model or made their data available. Our findings suggest a greater prominence of the general factor in all the indicators associated with the explained and reliable common variance, stability, and construct's replicability. In contrast, the specific factors related to item wording exhibited low values in the indices of replicability and consistency. These findings support the argument that the RSES is essentially a one-dimensional instrument. Most generally, the results question the argument that wording factors can be considered appropriate criteria for examining the underlying nature of wording effects.