Calibration and validation of the Body Self-Image Questionnaire using the Rasch analysis
MetadataShow full item record
The purpose of this study was to calibrate and validate the Body Self-Image Questionnaire using the Rasch analysis. The data from 1021 undergraduate students were used for this study. The Body Self-Image Questionnaire consists of 39 items under the nine factors related to the body image construct and a Likert-type five-point response scale for each item. The data from each subscale of the questionnaire were initially calibrated using the rating scale model for investigating category function and item structure. Violations in category function were found from the initial calibrations for the fatness evaluation (FE), social dependence (SD), height dissatisfaction (HD), and investment to ideals (II) subscales, and the optimal categorization was determined for those subscales. The collapsed four-point categorization obtained by combining categories three and four functioned better than other combinations for the FE, SD, HD, and II subscales and the original categorization was retained for the other subscales. Three misfitting items were also identified and deleted from corresponding subscales for further analysis. The revised categorization and item structure were cross-validated using a validation sample (n = 510) randomly selected from the total sample. Similar patterns of categorization were observed and confirmed except for the categorization for the HD subscale. Hierarchical orders of item difficulties for the validation sample were identical to the total sample. To Provide evidence of construct validity, three groups were formed based on body mass index (BMI) scores and the means in logits for the three BMI-based groups were compared and contrasted. Overall discrimination among groups for each subscale was effective. The result showed that the underweight BMI group tended to endorse categories indicating higher satisfaction with body image while the overweight BMI group tended to endorse categories indicating lower satisfaction with body image. The findings from these analyses supported that the data fitted the rating scale model well in terms of fit statistics, and the rating scale model adequately contrasted items and participants according to their measures in logits. The rating scale model provided a way to transform the ordinal data into interval and to investigate the category function of Body Self-Image Questionnaire.