Çoklu Göstergeler Çoklu Nedenler ve Lojistik Regresyon Yöntemlerinin Değişen Madde Fonksiyonunu Belirleme Performansları
Özet
In this thesis, DIF detection performances of MIMIC and LR methods for dichotomous data are investigated. Performances of these two methods are compared with according to their type-I error and power rates. Conditions covered in the study are: sample size (2000 and 4000 respondents), ability distributions of groups (N (0,1) and N( -0.5,1)), and the percentage of items with DIF (10% and 20%). Ability parameters of the respondents in the reference group (N(0,1)), exam length (30 items), the variation in item power parameters between groups for the items that contain DIF (0.6), and the ratio of focus group to reference group (1:1) are the conditions that are held constant. In the survey, performances of the methods are investigated with respect to 8 distinct simulation conditions. MPlus 6 and SAS 9.1.3 computer applications are used for MIMIC and LR DIF detection methods, respectively.