Examining the effect of different scoring methods on test and item properties in multiple choice tests

This study examined the effects of using number-right (1-0) scoring, experimental weighting, and confidence test scoring on a multiple-choice mathematics achievement test's item properties, validity, and reliability. The test comprised 20 items, each with 4 options, and was administered to 306 eighth-grade students from 5 primary schools in Istanbul's Kartal and Pendik districts during the 2023-2024 Fall semester. As a result of the research, the lowest average item difficulty index was observed with the first version of confidence test scoring, while the highest was with experimental weighting scoring. Item discrimination indexes were higher with both versions of confidence test scoring compared to number- right scoring. In the first version of confidence test scoring, 14 items showed higher discrimination indexes than the second version. For the validity analysis of the research, criterion validity was tried to be determined. Students' school academic achievement was accepted as criteria, and a positive and significant relationship was found between these scores and the scores obtained from all scoring methods used in the research. The highest correlation with criterion scores was seen in the second version of confidence test scoring, while the lowest was in experimental weighting scoring. According to the reliability analysis of the research, the KR-20 coefficients of the first version of the confidence test scoring and the number-right (1-0) scoring were found to be 0.87 and 0.81, respectively, and the Cronbach alpha coefficients of the second version of the confidence test scoring and the experimental weighting scoring were found to be 0.86.