JOURNAL OF MEASUREMENT AND EVALUATION IN EDUCATION AND PSYCHOLOGY-EPOD, vol.1, no.2, pp.78-84, 2010 (ESCI)
It is common to see hierarchical or nested data structure in many research areas. In education, students are placed within a hierarchical social structure such as classroom, school, city, country. The students in the same classroom or school exhibit more similar characteristics to each other than the students who are randomly selected from the population. Therefore it can not be said that the observations gathered from individuals in the same social unit are fully independent. However, one of the important assumptions in traditional statistical analysis procedures is the independence of observations. The problem that arouses with hierarchial data structures can be addressed with the hierarchical linear modeling procedures. In this paper, two-level hierarchical linear models have been introduced and the comparison of the analyses with simple linear regression and hierarchical linear modeling on TIMSS 1999 Turkey data have been provided.