Item Analysis for Examination Test in the Postgraduate Student's Selection with Classical Test Theory and Rasch Measurement Model

Abstract

University entrance exams are conducted to ensure applicants' qualifications are placed into the program of their choice. Test results have important and significant value in making the right decision about the suitability of the applicant; the validity of the exam is significant to achieve the objectives set. The purpose of this study is to provide empirical evidence of the validity of the new construct in developing the Academic and English Test Exams using the Classical Test Theory and the Rasch Measurement Model. Admission Test for postgraduate entrance examination consisting of 120 multiple choice items with five answers/option choices (A-E) and has been developed and assessed by experts who are competent in their fields and questions are given to 409 postgraduate entrance exam participants. Software applications used for CTT and Rasch Model are ITEMAN version 3 and JMETRIK version 4 windows, where the application is free of licenses. The software automatically generates parameter estimation recommendations for assessing the quality of test items. The CTT results identified 39 questionable items using difficulty and index discrimination. Rasch's results show that the statistics of people (Separation 2.55> 2.00 and reliability 0.87> 0.80) and item statistics (Separation 9.4> 3.0 and reliability 0.99> 0.8) are excellent person and item reliability. Overall, using the Rasch model obtained 68 constructs that incorrectly matched items, as well as irrelevant identified, are suggested to be removed. While CTT provides limited information on two parameters, Rasch's results provide very detailed information about the quality of the items being tested. Thus the two models can be integrated to produce sufficient evidence of validity and reliability items in the development of standardized tests. Even from the second approach, the model produced 28 items in common as problem items. These results indicate that more items are recommended for removal by the Rash model than the CTT can be linked to the procedure followed by two frameworks in determining the quality of test items.