Messick의 타당도 관점에서 Rasch 측정모형 적용을 통한 대학 강의평가 도구개발의 타당화

초록

This study investigates the questionnaires which is used to evaluate instruction in universities. Specifically, this study aims to test the validity of items within the framework of the content, substantive, structural, generalizability, external, and consequential aspects of validity as described by Messick(1995). This study is performed by the application of rating scale model which is based on the Rasch model. The results can be summarized as the four findings. First, fit statistics(INFIT and OUTFIT) for items were between 0.65～1.49, suggesting that items fit the Rasch measurement model. The point-measure correlation coefficient reflects the correlation between measurement of a single item and the Rasch measure of the scale. The point-measure correlation coefficient for all 15 items in this study were above 0.5, indicating the higher the relationship between the item value and the total score measure. Second, person fit statistics also revealed that approximately 11% in the study sample showed fit values greater than 1.5. This implies that the Rasch model requirements are not being severely violated to accommodate the data from this sample. Further, the category statistics from the Rasch analysis indicated that the average measures advance monotonically with category. Third, To investigate the dimensionality of the scale, Principal component analysis of standardized residuals was performed. The result indicated that the percentage of variance accounted for by the Rasch model was 71.1%. Fourth, the differential item functioning for gender indicated 1 out of 15 items as displaying DIF.

키워드