Investigation of measurement invariance of literacy tests in the programme for international student assessment (PISA-2012) Uluslararası öğrenci değerlendirme programı (PISA-2012) okuryazarlık testlerinin ölçme değişmezliğinin incelenmesi1


Alatlı B. K., Bökeoğlu Ö.

Elementary Education Online, cilt.17, sa.2, ss.1096-1115, 2018 (Scopus) identifier

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 17 Sayı: 2
  • Basım Tarihi: 2018
  • Doi Numarası: 10.17051/ilkonline.2018.419357
  • Dergi Adı: Elementary Education Online
  • Derginin Tarandığı İndeksler: Scopus, TR DİZİN (ULAKBİM)
  • Sayfa Sayıları: ss.1096-1115
  • Anahtar Kelimeler: Bias, Differential item functioning, Invariance of measurement, Literacy, PISA
  • Ankara Üniversitesi Adresli: Evet

Özet

© 2018, Ankara University. All rights reserved.In this study, it was aimed at examining the measurement invariance of mathematical and science literacy tests and subtests of reading skills of PISA – 2012 in terms of the language variable. The research was designed as a descriptive study. The study was conducted on the data obtained from the test booklet-3 of PISA 2012 (it is the only booklet that had the highest number of items published and also had items from all of the three parts) which was responded by 2149 students from Australia, France, Shanghai – China and Turkey. The invariance of factor structures was examined by Multi-Group Confirmatory Factor Analysis (MGCFA) while the differential item functioning (DIF) was examined by Hierarchical Generalized Linear Modeling (HGLM), Simultaneous Item Bias Test (SIBTEST) and Likelihood Ratio (LR) based on Item Response Theory (IRT) in relation to language forms of the tests. In addition to this, experts’ opinions were asked in order to determine the possible bias sources. According to MGCFA results, PISA – 2012 mathematical and science literacy and subtests of reading skills demonstrated structural invariance while there was no metric invariance. The items which demonstrated at least B level DIF according to all three DIF technique results were accepted as showing DIF. According to DIF analyses, the DIF items comprised of 35%, 34% and 22% of mathematical and science literacy tests and subtests of reading skills respectively while the items showing measurement invariance comprised of 24%, 7% and 36% of the tests in terms of language variable. The highest number of DIF items in all three parts was determined between the groups that took English and Chinese forms. As a result of bias analysis, some possible bias sources were as follows; using the words or phrases in different meanings in the items, being familiar to the item content in favor of a language group and the differentiation of the items depending on the translation.