碩士 / 國立中正大學 / 心理學研究所 / 92 / Studies of detecting differential item functioning (DIF) in polytomous items have gained much attention in recent years. This study compares two methods of detecting DIF in polytomous items under various simulated conditions: logistic discriminate function analysis (LDFA; Miller & Spray, 1993) and common slope multinomial logistic regression (MLR; French & Miller, 1996; Hidalgo-Montesinos & Gomez-Benito, 2003). Experiment 1 focuses on the graded response model (GRM) and Experiment 2 on the partial credit model (PCM). The dependent variable are empirical Type I error rate and statistical power. One hundred replications were made under each condition.
The results show that both methods performed better under the PCM than the GRM, when tests contained solely polytomous items than when tests contained both dichotomous and polytomous items, when the DIF patterns were nonuniform than they were uniform. When tests contained solely polytomous items, LDFA yielded slightly better results than MLR. When tests contained both dichotomous and polytomous items, they performed similarly. Therefore, LDFA is recommended when tests contain solely polytomous items (e.g., Likert-type or rating scale items) and both methods are recommended when tests contain both dichotomous and polytomous items.
Identifer | oai:union.ndltd.org:TW/092CCU00071027 |
Date | January 2004 |
Creators | 鄭致寯 |
Contributors | Wen-Chung Wang, 王文中 |
Source Sets | National Digital Library of Theses and Dissertations in Taiwan |
Language | zh-TW |
Detected Language | English |
Type | 學位論文 ; thesis |
Format | 144 |
Page generated in 0.0015 seconds