Kayıp Veri ile Baş Etme Yöntemlerinin Ölçme Değişmezliğine Etkisi Açısından Karşılaştırılması

Abstract

The purpose of this study is to compare the missing data handling methods in terms of the influence on measurement invariance. For this purpose, data of 5496 students participated in PISA 2015 tests from Turkey and responded to the items about science learning motivation was used. Sampling of 300, 1000, and 2000 students were created from the complete data to research on missing data of different sample sizes and different rates. 5%, 10% and 20% missing data were generated as having missing completely at random (MCAR) mechanism from the size of each sample. In all data sets, missing data was completed with listwise deletion (LD), serial mean imputation (SMI), regression imputation (RI), expectation maximization (EM) and multiple imputation (MI) methods, and measurement invariance study between genders was investigated with multiple-group confirmatory factor analysis. Findings from each dataset were compared with reference values. In the result of the study, EM method in the data set with 300 students and 5% missing, MI and EM methods in the data sets with 300 students and 10% and 20% missing gave more similar results than the other methods. In data sets of 1000 students, RI method in the data set with 5% missing, EM and MI methods in the data set with 10% missing, EM method in the data set with 20% missing show more similar results than the other methods. In the data sets of 2000 students, RI, EM and MI methods in the data set with 5% missing, MI method in the data set with 10% missing and EM method in the data set with 20% missing gave more similar results than the other methods.