Factor Structure and Incremental Validity of the Original and Modified Versions of the Difficulties in Emotion Regulation Scale

Abstract

The Difficulties in Emotion Regulation Scale (DERS; Gratz & Roemer, 2004) is a self-report measure that assesses six facets of emotion dysregulation. A modified version of the DERS (M–DERS) was developed to address psychometric limitations of the original measure (Bardeen, Fergus, Hannan, & Orcutt, 2016). Although the factor structure of the M–DERS (i.e., two models: correlated trait and second-order models) has been supported via confirmatory factor analysis (CFA), the tenability of a bifactor model of the M–DERS has yet to be examined. Preliminary research suggests that a bifactor model of the M–DERS is tenable. In this study (Ns of 993 and 578), results from a series of CFAs indicated adequate fit of the M–DERS and poor fit of the original DERS across several tested models (e.g., correlated trait, second-order, bifactor). Although a considerable amount of variance was accounted for by the general factor, statistical indexes from the bifactor model supported a multidimensional conceptualization of the M–DERS. The Nonacceptance and Goals subscales evidenced incremental utility, after accounting for the general factor, in predicting general distress (Nonacceptance only) and intolerance of uncertainty. Implications for future use of the DERS and M–DERS are discussed.

Notes

1 Goodness-of-fit statistics, using adjusted weighted least squares estimation rather than robust maximum likelihood estimation to test all models, are presented in Table S1.

2 Post-hoc EFAs were conducted, per reviewer request, on the DERS (N = 825) and M–DERS (N = 564) using parallel analysis with principal axis factoring (O’Connor, 2000). An uninterpretable nine-factor solution was identified via parallel analysis for the DERS. The first to second eigenvalue ratio was 14.89 to 3.71. Consistent with previous research (Bardeen et al., 2016), a five-factor solution was identified for the M–DERS. The first to second eigenvalue ratio was 15.63 to 2.60. Factor 1 accounted for 53.95% of variance, and 19.66% of variance was attributed to the remaining four factors. These results provide additional support for the assertion that the M–DERS is a multidimensional measure with a strong general factor.

3 Structural regression models and parameter estimates are depicted in Figures S1 and S2. The standardized factor loadings for the manifest indicators in both of these structural regressions are presented in Tables S2 and S3.