MANTEL-HAENSZEL ANALYSIS OF GENDER DIFFERENTIAL ITEM FUNCTIONING (DIF) IN MALAWI SCHOOL CERTIFICATE OF EDUCATION (MSCE) COMPUTER STUDIES EXAMINATION ITEMS

Abstract

The purpose of this study was to re-examine gender Differential Item Functioning (DIF) in Malawi School Certificate of Education (MSCE) Computer Studies examination items using a Classical Test Theory (CTT)-based Mantel-Haenszel (MH) DIF detection procedure. MSCE Computer Studies examinations are CTT-based. Analysis of DIF is there to enhance the validity of examinations. The study was a partial replication of a research study by Jailosi (2012) who used Item Response Theory-Likelihood Ratio (IRT-LR) DIF detection procedure on a CTT- based examination. IRT procedures have very stringent assumptions that are practically very difficult to meet. One of the assumptions is that there must be model data fit where even a slight model misfit to the data can be mistaken for DIF (Teresi, 2004). The current study was undertaken to find out the extent to which MH DIF detection procedure could detect DIF in MSCE Computer Studies examination items; determine the group which the detected DIF disadvantaged and establish the type of DIF that was detected. Thirty-four items were analysed for a random sample of 500 examinees (300 males and 200 females). Data was analysed using Generalized Mantel-Haenszel DIF (GMHDIF) Program by Fidalgo (2011). Results have shown that category „B‟ and „C‟ DIF magnitude was detected in 3 out of 34 items representing 8.82%. Out of the detected items, 66.67% had uniform DIF whereas 33.33% had non-uniform DIF. Both items with uniform DIF disadvantaged females. Results were discussed concerning the appropriateness of the DIF detection method based on the nature of the data. It was recommended that the examining board consider adopting gender DIF analysis to enhance examination quality.