Abstract
The current study compares the psychometric properties of the well-known Physical Self-Description Questionnaire (PSDQ; Marsh, Richards, Johnson, Roche, & Tremayne, 1994) with respect to different number of response categories. Mixture Distribution Rasch models (Rost, 1990) are applied to PSDQ data obtained with two rating scale variants: The original rating scale with 6 response categories, and a variant with 4 response categories. Data are collected from two samples consisting of N = 806 (original rating scales with 6 response categories) and N = 905 (4 response categories) youth from age 12 to 18.
In most cases, 2 class solutions fit the data best, suggesting the presence of Differential Item Functioning respectively response sets. Logistic regression analyses help identify the characteristics of respondents in either of the 2 latent classes. In particular, youth with lower educational levels appear to have problems with using the rating scales correctly.
The results suggest that the administration of the PSDQ using a rating scale with as many as 6 response categories can be problematic because the test scores’ psychometric qualities may be different across subgroups. In contrast, using only 4 response categories leads to improved psychometric quality. We also discuss implications for the use and development of assessment measures in general.
