Investigating Psychometric Properties of the Self-Compassion Scale Using Rasch Methodology

Abstract

The 26-item Self-Compassion Scale (SCS) and its short 12-item version (SCS-SF) were reported to have acceptable psychometric properties, and both scales are widely used to assess self-compassion in individuals. However, recent investigations were inconsistent regarding factor structure of the SCS, and psychometric properties of the scale were not tested for consistency with principles of fundamental measurement using appropriate methodology such as Rasch analysis. A partial credit Rasch model was used to evaluate psychometric properties of the SCS and SCS-SF with the sample of 743 respondents randomly divided into two equal subsamples (A and B) to replicate the results for the purpose of robustness. Initially, there were no misfitting items but the local dependency between various items affected Rasch model fit. This issue was resolved by combining locally dependent items into four super-items resulting in the best fit to the Rasch model of both SCS and SCS-SF, with evidence of unidimensionality and an excellent sample targeting. Although both scale versions had strong reliability satisfactory for individual and group assessment, the original SCS demonstrated superior psychometric properties reflected by higher reliability indicated by Person Separation Index (PSI) of 0.90 compared to the SCS-SF (PSI = 0.85). These analyses were replicated with the sample B for both scale versions, resulting in equally good fit. This permitted generating ordinal-to-interval conversion tables based on Rasch model estimates. The current study supported the reliability and internal validity of both the SCS and SCS-SF. Accuracy of these assessment instruments can be further improved by using the ordinal-to-interval conversion tables published here.