Differential item functioning in the figure classification test

Abstract

The elimination of unfair discrimination and cultural bias of any kind, is a contentious workplace issue in contemporary South Africa. To ensure fairness in testing, psychometric instruments are subjected to empirical investigations for the detection of possible bias that could lead to selection decisions constituting unfair discrimination. This study was conducted to explore the possible existence of differential item functioning (DIF), or potential bias, in the Figure Classification Test (A121) by means of the Mantel-Haenszel chi-square technique. The sample consisted of 498 men at a production company in the Western Cape. Although statistical analysis revealed significant differences between the mean test scores of three racial groups on the test, very few items were identified as having statistically significant DIF. The possibility is discussed that, despite the presence of some DIF, the differences between the means may not be due to the measuring instrument itself being biased/ but rather to extraneous sources of variation, such as the unequal education and socio-economic backgrounds of the racial groups. It was concluded that there is very little evidence of item bias in the test.

Opsomming
Die uitskakeling van onregverdige diskriminasie en kultuursydigheid van enige aard, is tans 'n omstrede kwessie in die werkpiek in Suid-Afrika. Ten einde regverdigheid in toetsing te verseker, word psigomefrriese toetse onderwerp aan empiriese ondersoeke na die moontlikheid van sydigheid wat kan lei tot keuringsbesluite wat onregverdige diskriminasie meebring. Hierdie ondersoek is ondemeem om die moontlikheid van differensiele itemfunksionering (DIF), of potensiële sydigheid, in die Figuurindelingtoets (A121), met behulp van die Mantel-Haenszel chikwadraattegniek, te ondersoek. Die steekproef het bestaan uit 498 mans by 'n produksiemaatskappy in die Wes-Kaap. Alhoewel statistiese ontleding beduidende verskille in gemiddelde toetstellings van drie rassegroepe op die toets aangedui het, is bate min items aangedui wat statistics beduidende DIF bevat. Die moontlikheid word bespreek dat, hoewel sommige DIF in die toets teenwoordig is, die verskille tussen die gemiddeldes nie die gevolg is van 'n sydige meetinstrument per se nie, maar eerder die gevolg van eksteme bronne van variasie, soos byvoorbeeld die ongelyke opvoedkundige- en sosio-ekonomiese agtergronde van die rassegroepe. Die gevolgtrekking was dat daar bate min getuienis van itemsydigheid in die toets is.