An examination of the psychometric properties of the physical self-description questionnaire using a polytomous item response model

    loading  Checking for direct PDF access through Ovid


Objectives:This study proposes polytomous item response theory (IRT) as a method for item analysis of Likert-type responses that are commonly used in sport and exercise psychology measures. Thus, the aims of this study were to determine the psychometric qualities of the items at the endorsement option, item and test level.Method:The graded response IRT model (Samejima, Psychometrika Monograph 17 (1969)) was fitted to the 70-item Physical Self-Description Questionnaire (PSDQ) (Marsh et al., J Sport Exer Psychol 15 (1994) 270), which was administered to a large sample of Australian adolescents enrolled in a metropolitan sports high school.Results:By analyzing in greater detail the psychometric properties of items, the results show how polytomous IRT can provide new insight into item analysis that is not available in traditional classical test analyses. The graded response IRT model (Samejima, Psychometrika Monograph 17 (1969)) allowed for the identification of items that provided high and low measurement precision, items that needed rewording, and items that are redundant in that they add little information or had redundant response options.Conclusion:The PSDQ was most discriminating among participants with lower estimates of physical self-concept. Furthermore, the study showed how choosing items to maximize reliability may not always be the optimal strategy. IRT models provide item/sample free calibration, local standard errors, and give more information at the item level by offering additional insights as to qualities of the items over and above those gleaned from the classical test theory models (CTT). The increased information at each Likert-scale point, and the added information from the polytomous IRT model, make this an attractive approach for the development of psychological measures using a polytomously scored item format.

    loading  Loading Related Articles