Elif Ozlem ARDIC1 Selahattin GELBAL2
1Hacettepe University, TURKEY
2Hacettepe University, TURKEY
DOI: 10.14689/ejer.2017.68.12
ABSTRACT
Purpose: The aim of this study was to examine measurement invariance of the interest and motivation related items contained in the PISA 2012 student survey with regard to gender school type and statistical regions and to identify the items that show differential item functioning (DIF) across groups.
Research Methods: Multiple-group confirmatory factor analysis was conducted to examine measurement invariance. When the invariance with regard to gender was being investigated, potential item biases were examined, as the criteria used in the model fit evaluation were not met. Mantel-Haenszel, poly-SIBTEST, and item response theory likelihood ratio (IRT-LR) techniques were employed to identify which items displayed DIF.
Findings: Results of the invariance test conducted based on the school type and statistical regions demonstrated that the models satisfied all invariance conditions. Failure to achieve measurement invariance according to gender indicates that at least one of the items in the scale displayed DIF. When the results of DIF according to gender were examined, MH identified DIF in six items at A level, poly-SIBTEST identified DIF in one item at A level, two items at B level, and three items at C level, IRT-LR identified DIF in two items at C level.
Implications for Research and Practice: Further studies could determine which techniques would be more suitable for which situations by conducting simulation studies along with real data, and explore the possible reasons why the items display DIF.
Keywords: PISA measurement invariance multiple-group confirmatory factor analysis, differential item functioning.