Akademik Veri Yönetim Sistemi
HACETTEPE UNIVERSITESI EGITIM FAKULTESI DERGISI-HACETTEPE UNIVERSITY JOURNAL OF EDUCATION, cilt.36, sa.2, ss.393-405, 2021 (ESCI)
is aimed in this study to identify whether the items in the Turkish subtest of the Transition from Basic Education to Secondary Education (TEOG) exam are biased. In this regard, whether the items show differential item functioning (DIF) is analyzed primarily according to the Mixed Item Response Theory (MixIRT) model and the Mantel-Haenszel (MH) method. The items that are determined to display DIF according to both methods are analyzed with regards to the subject field and cognitive level, as well as taking expert opinion on including bias. The sample of this study consists of 5000 students who were randomly selected from among the students that took the TEOG 2015 Turkish subtest. Analyses have shown that among the models based on MixIRT, the model with two latent classes is determined to fit the data best. When the items that indicate DIF according to both methods are analyzed in the DIF examination conducted based on the two latent classes that emerged according to the MixIRT, it is observed that four items that are all advantageous for the first latent group, are understanding level questions that contain the subjects of meaning in sentences and paragraphs. The other three items in favor of the second latent class were seen to be knowledge level questions measuring the Turkish grammar subject. In this context, it can be submitted that item features are favored for some groups and disadvantageous for some others.