Multi-Faceted Constructs in Abnormal Psychology: Implications of the Bifactor S - 1 Model for Individual Clinical Assessment

Abstract

Burns et al. (this issue) have shown that the application of the symmetrical bifactor model to attention-deficit/hyperactivity disorder (ADHD) and oppositional defiant disorder (ODD) symptoms leads to anomalous and inconsistent results across different rater groups. In contrast to the symmetrical bifactor model, applications of the bifactor S-1 model showed consistent and theoretically well-founded results. The implications of the bifactor S-1 model for individual clinical assessment are discussed. It is shown that individual factor scores of the bifactor S-1 model reveal important information about the profile of individual symptoms that is not captured by factor scores of the multidimensional model with correlated first-order factors. It is argued that for individual clinical assessment factor scores from both types of model (multidimensional model with correlated first-order factors, bifactor S -1 model) should be estimated and compared. Finally, a general strategy for choosing an appropriate model for analyzing multi-faceted constructs is presented that compares areas of applications for (1) the multidimensional model with correlated first-order factors, (2) the bifactor S-1 model with a general reference factor, and (3) the bifactor S - 1 model with a directly assessed general factor.

Keywords: Bifactor S-1 model; Bifactor model; Clinical assessment.

Fig. 1

Individual factor scores for two individuals according to (a) the bifactor S-1 model presented by Burns et al. (this issue) with hyperactivity-impulsivity (HI) as general reference factor (GRF-HI) and two specific reference factors for inattention (SRF-IN) and oppositional defiant disorder (SRF-OD) (see Fig. 2b), and (b) a multidimensional model with correlated first-order facet-specific factors (see Fig. 2a) for hyperactivity-impulsivity (HI), inattention (IN), and oppositional defiant disorder (OD). All factors are centered (mean of 0). The factor scores of the model in the multidimensional model are calculated based on the regression eqs. IN = 0.6 ·GRF-HI + SRF-IN and OD = 0.6 ·GRF-HI + SRF-OD. The regression coefficient of 0.6 was roughly based on the results in Burns et al. (this issue)

Fig. 2

Multidimensional models for analyzing multi-faceted constructs. (a) Multidimensional model with correlated first-order facet-specific factors for hyperactivity-impulsivity (HI), inattention (IN), and oppositional defiant disorder (OD). (b) Bifactor S-1 model presented by Burns et al. (this issue) with hyperactivity-impulsivity (HI) as general reference factor (GRF-HI) and two specific reference factors for inattention (SRF-IN) and oppositional defiant disorder (SRF-OD). (c) Bifactor S-1 model with a directly assessed general hyperactivity-impulsivity factor (GF-HI) and three specific reference factors for hyperactivity-impulsivity assessed in specific situations (SRF-HI1, SRF-HI2, SRF-HI3). Y_ij: observed variables, E_ij: error variables, λ_ij: factor loadings, i: indicator, j: facet

Fig. 3

Decision flow chart for selecting an appropriate model for analyzing multi-faceted constructs