| Summary: | 碩士 === 國立臺南大學 === 測驗統計研究所碩士班 === 97 === The purpose of this study was to investigate the impact of missing data on the detection of DIF items as well as the effects of different imputation methods. The study manipulated four missing data treatments (listwise deletion, zero imputation, two-way imputation, multiple imputation), two methods of DIF detection (Mantel-Haenszel , SIBTEST) under three missing rates (10%, 20%, 30%) and two DIF magnitude (0.5, 0.8) to examine the Type I error and statistical power of DIF detection. The sample size was fixed to 1000 for both focal and reference groups.
Comparing with other three imputation methods, the listwise deletion resulted lower Type I error as well as statistical power. It showed that missing data would have some impact on DIF detection. The reduced Type I error and statistical power can be recovered by imputing the missing data. There is no much difference in three imputation.
|
|---|
