Elementary Education Online, 13(3), 934-953, 2014. İlköğretim Online, 13(3), 934-953, 2014. [Online]: http://ilkogretim-online.org.tr Investigation of Placement Test in terms of Item Biasness Hülya KELECİOĞLU 1 Burçin KARABAY 2 Ersoy KARABAY 3 ABSTRACT. This study was conducted to analyze the Centralized High School Entrance placement test (SBS) in terms of item biasness. In this research, it is examined that whether Turkish, Science and Technology, Math and Social Sciences sub-tests items in 8 th grade placement test (SBS) applied in 2009 to study the biasness in terms of gender and school type. This research was conducted on randomly selected 7000 students from public schools, 1000 students from private schools and 1000 students from regional boarding primary education schools (YIBOs). The placement test (SBS) consists of Turkish, Math, Science and Technology, Social Sciences and Foreign Language subtests. In this study, the data of Foreign Language subtest wasn’t used. Turkish subtest consists of 23 items; Math, Science and Technology and Social Sciences subtests consists of 20 items each. To check whether the data prove the unidimensionality assumption, the factor analysis is applied to the items constructed in the sub- tests. At the end of the factor analysis, DIF analysis was maintained for 18 items in Turkish subtest, 14 items in Math subtest, 14 items in Science and Technology subtest and 18 items in Social Sciences subtest (64 items in total). It was investigated whether there is DIF in the items of subtests or not with Mantel-Haenszel, SIBTEST and logistic regression methods. Items showing DIF according to at least two of these methods are handled as biased items and the reasons of biasness in these items are examined by taking experts’ opinions. All but 4 of items showing DIF were biased by experts. According to experts items may show bias because of the group properties and items. Experts could not give common and consistent explanations for three items in the Math subtest and one item of Turkish subtest showing DIF. It was concluded that resulting DIF in these items is not due to bias. Keywords: Bias, Placement Test, School Type, Gender, Sub Tests SUMMARY Purpose and Significance. In this research it is examined that whether Turkish, Science and Technology, Math and Social Sciences sub-tests items in the Centralized High School Entrance placement test (SBS) applied in 2009 to show biasness in terms of gender and school type. School type variable is limited to three categories: public school, private school and regional boarding primary education schools (YIBOs). In data analysis, DMF determination methods Mantel-Haenszel, logistic regression and SIBTEST are used. Items showing DMF are examined in the sense of biasness via at least two of the three methods. Item bias is a condition in which the possibility of one group to respond to an item correctly is higher or lower than another group is described as high or low possibility of responding an item correctly for one group (Zumbo, 1999). The possibility of responding to biased items correctly depends on a group of students achieve more than measured ability level of their group. This situation implies that evaluation tool makes an inaccurate measurement (Osterlind, 1983). Biasness, in a sense, shows that measurement consists of systematic error due to the fact that the direction and amount of bias changes according to the groups (i.e. gender, school type, socio-economic level, etc.) that individuals belong to. Method. The participant group of the study consists of students throughout Turkey participated in SBS in 2009. In the group of 8 th grade students who answered the questions on Booklet A in 2009 SBS, 470.518 of them studied in public schools, 19.369 of them studied in YIBOs and 15.875 of them studied in private schools. For this research, 7000 students from public schools, 1000 students from private schools and 1000 students from YIBOs are randomly selected. 262.902 of the students who took Booklet A in 2009 SBS were male and 242.862 of them were female. For this research, 7000 male and 7000 female students were randomly selected. In this research, two data sources were used; first, 8 th grade students’ responses to 4 subtests in Booklet A of SBS (2009) and second, a questionnaire to take experts’ opinion for indicating why items containing DMF show biasness. Responds that students gave to the subtests in SBS (2009) were taken from National Ministry of Education (NME), Directorate of General of Educational Technologies. Items found DMF according to at least two methods were broached to item writers of Turkish, Math, Science and Technology and Social Sciences, and to teachers working in primary schools. To check whether the data prove the unidimensionality assumption, the factor analysis is applied to items constructed in the sub-tests. At the end of the factor analysis, DMF analysis was conducted for 18 items in Turkish subtest, 14 items in Math subtest, 14 items in Science and Technology subtest and 18 items in Social Sciences subtest (totally for 64 items). In this study, Mantel-Haenszel, SIBTEST and logistic regression analysis are done via three different methods to determine whether items figured in Turkish, Math, Social Sciences and Science and 1 Professor Hülya Kelecioğlu. Hacettepe University Measurement and Evaluation Department, hulyakelecioglu@gmail.com. 2 Expert Burçin Karabay. Ministry of National Education General Directorate of Lifelong Learning, Beşevler/Ankara, batanburcin@yahoo.com. 3 Expert Ersoy Karabay. Ministry of National Education General Directorate of Innovation and Educational Technologies, Beşevler/Ankara, ekarabay@yahoo.com.