| Summary: | 碩士 === 國立屏東教育大學 === 教育心理與輔導學系碩士班 === 102 === The purpose of this study is to examine item differential functioning (DIF) of the Self-Description Questionnaire (SDQ-I) using Rasch model. The participants were recruited 474 children from several elementary schools. In examination on dimensionality of the SDQ-I, the resutls that items associated with Parent-relations factor in Non-Academic subscale exhibited unidimensionally. Thus, DIF analyses were conducted individually based on four subscales (i.e., Academic self-concept, Non-Academic self-concept subscales, General self-esteem, and Parent-relations factors). The main results in the study were delineated as follows:
DIF Analyses across Gender and Grades
1.There were 16 (53%)-17 (57%), 8 (22%)-11 (31%), 2 (20%), and 1 (11%) gender DIF items in Academic, Non-Academic subscales, General, and Parent factors, respectively. Items associated with Math and Psychical self-concept favored the boys; items regarding Reading, Appearance, and Peer-relations self-concept favored the girls. Specifically, these DIF items could be explained by gender stereotypes.
2.There were 2 (7%)-3 (10%), 10 (28%), 1 (10%), and 1 (11%)-5 (55%) grade DIF items in Academic, Non-Academic subscales, General, and Parent factors, respectively. Items associated with Math and Appearance self-concept favored the lower-grade students; items regarding Reading and Peer-relations favored the higher-grade students.
Gender and Grade Differences on SDQ-I
1.Comparing gender differences on each subscale, the results found boys tended to have higher scores on Academic subscale at time1 and Non-Academic subscale than girls did. However, no gender differences were found in General, and Parent factors. Regarding grade differences, no significant differences were identified in Academic, Non-Academic subscales, General, and Parent factors.
2.When deleting DIF items, gender differences were found in Non-Academic subscale, suggesting that boys tended to obtain higher scores than girls did. However, no gender differences were found in Academic subscale, General, and Parent factors. Regarding grade differences, no grade differences were identified in Academic, Non-Academic subscales, General, and Parent factors.
The Effects of DIF on Gender and Grade Differences
1.The effects of DIF on gender differences were found in Academic and Non-Academic subscales; however, there were inconsequential effects of DIF on grade differences in each subscale.
DIF Detection for Longitudinal Analyses
1.There were 6 (20%), 2 (6%), 2 (20%), and 1 (11%) DIF items across different points in Academic, Non-Academic subscales, General, and Parent factor, respectively. Also, non-invariant thresholds in SDQ-I subscales were identified.
2.When adjusting DIF items, there were no significant differences across time in Academic, Non-Academic subscales, General, and Parent factor. In contrast, without adjusting DIF items, there were significant differences across time in Non-Academic subscale, General, and Parent factors. As a result, when there are no DIF items of SDQ-I across time, can the differences of self-concept across pretest and posttest be justified.
Finally, the study proposed some promising suggestions for further studies as well as practitioners.
|
|---|
