Should We Use a Total Score, Two Subscale Scores, or Six Subscale Scores for the Self-Compassion Scale? A Multi-faceted Assessment Beyond Model Fit Indices

Abstract

Objectives: The aim of this study was to conduct a multi-faceted assessment of the psychometric properties of the Self-Compassion Scale (SCS). In addition to the goodness-of-fit, we aimed to assess the strength and replicability of three factorial solutions, and the quality and effectiveness of the three scoring schemes of the scale (i.e., total scale score, two subscale scores, and six subscale scores). Methods: Participants were 1508 Spanish-speaking community-dwelling adults (M = 34.94 years, SD = 15.02). Data were examined by means of a conjoint strategy using Rasch modeling, non-linear factor analysis, exploratory bifactor analysis, and parallel analysis. A procedure for selecting the optimal set of items that must be used to compute individual’s scores was used. Results: The unidimensional solution showed a marginal model fit (RMSR =.089), and both the bifactor two-group and bifactor six-group solutions showed a good fit (RMSR =.043 and.019, respectively). However, only the unidimensional and the bifactor two-factor solutions showed interpretable and replicable factor structures, and high-quality and effective scores to be used for measurement purposes. Subscale scores derived from the six primary factors did not show adequate psychometric properties. It was observed that the information provided by 10 items was redundant and had already been provided by the other 16 items. Conclusions: Good model fit is neither sufficient nor necessary to justify the use of a scoring scheme. Goodness-of-fit statistics should be complemented by an assessment of the metric properties of the resulting scores when proposing SCS scoring schemes.