Analisis Aitem Subtes IN 2 (Unfitting Numbers) dan Aitem Subtes DN 2 (Arithmetic Reasoning) pada USU General Reasoning Test (U-GRT)

Abstract

This study used a descriptive quantitative method using a sampling technique, namely convenience sampling. The purpose of this study was to analyze the IN 2 subtest item (Unfitting Numbers) and the DN 2 subtest item (Arithmetic Reasoning) on the USU General Reasoning Test (U-GRT). Analysis of the 40 items of the IN 2 subtest and 37 items of the DN 2 subtest was carried out using Rasch modeling which included analysis of the reliability coefficient, analysis of the validity coefficient, and analysis of the item difficulty index. This study uses secondary data that has been collected by previous researchers who have been carried out by giving classical tests to 207 samples with an age range of 15-25 years divided into 4 sets (A, B, C, D). The reliability coefficient uses person reliability in the IN 2 subtest which produces a value of -0.126 in set A, 0.607 in set B, 0.279 in set C, and 0.424 in set D which are categorized as weak for all sets. The validity coefficient using the MNSQ infit value and the MNSQ outfit has a value in the range of 0.5-1.5 in 38 items so that it is stated well in the measurement. However, there were 2 items that were invalid, namely items 13 and 15. The item measure values obtained ranged from -1,507 to 2,833. The reliability coefficient using person reliability in the DN 2 subtest yields a value of 0.218 in set A, 0.285 in set B, 0.495 in set C and 0.710 in set D so that sets A, B and C are categorized as weak and in set D are categorized as sufficient. The validity coefficient using the MNSQ infit value and the MNSQ outfit has a value in the range of 0.5-1.5 in 36 items so that it is stated well in the measurement. However, there is 1 invalid item, namely item 29. The item measure values obtained range from -0.891 to 2.320. The results of the research on the two subtests show that the validity of the dominant test is good, but elimination of invalid items in both subtests. In addition, it is necessary to re-test with a larger number of samples in order to obtain better reliability and item difficulty index results.