The Effect of Missing Data Handling Methods on Differential Item Functioning with Testlet Data

This study examined the effect of three missing data handling methods (listwise deletion, zero imputation and fractional hot-deck imputation) on differential item functioning (DIF) with testlet data with a variety of sample size and missing data percentage under missing completely at random , missing at random, and missing not at random missing mechanisms. The study was conducted on two different datasets consisting of six testlets which contain 20 reading comprehension items of a foreign language test. Data with left-skewed distribution was referred to as data1 and data with right-skewed distribution was referred to as data2. In current study false DIF was identified in data1 with all missing data methods under the missing at random mechanism with a 5% missing data rate in small sample size. Similarly, in analyses performed under the missing at random mechanism for data2, the proportion of items classified as false DIF was notably higher in the small sample size. Results also indicated that in all conditions, list wise deletion had the lowest correlations with DIF values obtained from the original datasets, datasets containing no missing data and serve as a reference for comparative analyses with datasets where missing data were artificially introduced. The zero imputation and fractional hot-deck imputation methods produced similar correlations when the missing data percentage was set at 5%. However, in the case of 15% missing data, zero imputation exhibited higher correlation values. Besides, in all conditions correlation values decreased with the increase of missing data percentage regardless of the missing data handling method.