THE FACTORS EXPLAINING THE DIFFERENTIAL ITEM FUNCTIONING IN THE ADMINISTRATION OF TIMSS 2011 SCIENCE TEST ACCORDING TO GENDER

Dr. Seher Yalcin
Dr. Ezel Tavsancil

Ankara University

Abstract

The objective of this study is to define differential item functioning (DIF) according to the gender and if DIF is determined in the items posd to measure science literacy in TIMSS (Trends in International Mathematics and Science Study) 2011, the role of the students' and schools' characteristics as the sources of DIF in the determined items are analyzed with the frame of multilevel models. Also defining the possible sources of DIF in the items with the judgmental approach is aimed with this study. The whole eight grade students in Turkey are composed of the population of this descriptive study. The sample of this study was consisted of application in TIMSS 2011. TIMSS sampling which it was defined by stratifying according to the geographical regions and the types of schools that attended the TIMSS 2011 and comprised of 239 schools and 6928 eight grade students. In all of the items, DIF is determined through full bayesian estimation in multilevel modeling approach. Analyzes of DIF are progressed by WinBUGS for all booklets. At the end of the study, it is determined that the distribution of 39 items in which DIF is detected is equal by sex. The variables, which are examined in student level, explain DIF in two items. These items displaying DIF in favor of boys don't display DIF anymore. The variable of "self-confidence regarding to science" in student level explains differential item functioning in maximum number of items. It is seen that the variable of "school sources", which is examined in school level, is related with some items displaying DIF by sex. It is determined that while the judgments of experts are consistent with the statistical results in some of the items, they are rather different in another items. Besides, it is detected that there is difference of opinion among experts in some of the items.

Keywords

Differential item functioning, Multilevel models, Science, TIMSS 2011.

FULL TEXT