INVESTIGATION OF THE EFFECT OF MISSING DATA ON DIFFERANTIAL ITEM FUNCTIONING IN MIXED TYPE TESTS

In this research, the aim was to examine effects of Markov Chain Monte Carlo (MCMC), multiple imputation (MI), and expectation maximization (EM), all methods coping with missing data in mixed type tests including dichotomous polytomous items, on differential item functioning (DIF). The study carried out a complete set consisting scores 1160 students who took booklet number 9 science test Trends International Mathematics Science Study (TIMSS) 2019 answered it full. conditions be examined for effectiveness were mechanism (MCAR MAR), DIF level (A, B, C), rate (10% 20%). Data assigned sets created by deleting at different rates under completely random (MCAR) (MAR) mechanisms over aforementioned set. analysis performed obtained poly-SIBTEST method using MCMC, MI, EM methods. results from then compared result implications other reference. showed that MCMC better C-level than A B levels terms examined. MI observed most successful determining items showing 10% 20% MCAR mechanisms. When set, three similar MAR while gave closest mechanism.