Using mixture distribution models to test the construct validity of the Physical Self-Description Questionnaire

Tietjens M., Freund P., Büsch D., Strauss B.

Abstract

Objectives: The Physical Self-Description Questionnaire (PSDQ; Marsh, Richards, Johnson, Roche, & Tremayne, 1994) is an internationally established instrument for the measurement of the physical self, featuring a total of 11 scales. So far, assessments of the psychometric quality of the PSDQ have applied CFA and MTMM analyses. In these assessments, some PSDQ scales have been found to be subject to, among others, a gender bias, and a bias toward negatively coded items. Probabilistic mixture distribution models such as the Mixed Rasch model (MRM; Rost, 1990) have not been used so far to test the psychometric qualities of PSDQ scores. Design and Method: Data are collected from German university sport and psychology students (N= 605, 270 males and 335 females). Both CFA and MRM analyses are conducted. Results: CFA results show acceptable goodness-of-fit indices for the 11-factor solution. The majority of the 11 MRM analyses indicate that the PSDQ scales can be considered unidimensional in the sense of a Rasch model for polytomous data. However, the scales physical activity, physical self-concept, and strength reveal shortcomings. Conclusion: For a rather large proportion of yet undefined persons the PSDQ may be problematic for diagnostic purposes, and researchers and practitioners should be aware of this. Future studies should also consider MRM analyses with samples from different age categories and manifest groups, which can be tested against statistically established latent groups. © 2012 Elsevier Ltd.