تدريج الاختبارات مختلطة الفقرات باستخدام نظرية الاستجابة للمفردة : مقارنة بين النماذج الأحادية ومتعددة الأبعاد
There is approximately full agreement among experts in measurement and evaluation that both multiple choice questions and assay questions have strengths and weaknesses and combining both types in single test increases the test accuracy in measuring the target trait from the logic that each type complements the other. However, combining both types into single test raise several questions one of them is these questions measure the same or similar ability or abilities or each type of measure different ability from the abilities are measured by the other types of items? based on this question, another question was raised: are the unidimensional item response modes valid to calibrate this type of test or these types of test require multidimensional item response theory models. Therefore, This study aimed to compare the relative accuracy of the unidimensional and multidimensional models in calibrating mixed item format tests. To achieve this gool, a test for measuring the math achievement among the six grade students and the test contains 15 questions dividing into 10 multiple choice questions and five assay questions; the sample consists of 738 students from the six grade primary schools, 400 female students, 338 male students. The two-parameter logistic model was used to calibrate the multiple choice questions and generalized partial credit model was used to calibrate the open ended questions using Parscale, and Mirt package from R was used to calibrate the multidimensional models. The results indicated that multidimensional models outperform the unidimensional models based on test information function and item goodness of fit for the data. (Published abstract)