Comparative Test of Cronbach's Alpha Reliability Coefficient, Kr-20, Kr-21, And Split-Half Method

Abstract

Final semester exams have an essential role in measuring student learning outcomes. Therefore, the tests must be of good quality. The problem is that lecturers must use reliable measuring tools to measure learning outcomes. This research aims to compare various reliability coefficients to determine which method has the highest reliability. The data collection methods in this research are observation, interviews, and tests based on student responses to the final semester exams. The data analysis technique uses the reliability of the Alpha Cronbach formula, Kuder Richardson 20, 21, and the split-half method consisting of Spearman-Brown, Flanagan, Rulon, and Hoyt. This type of research uses quantitative description. Comparative research designs are ex post facto. The research subjects were 90 undergraduate students. The research results provide several conclusions: first, the odd-even Rulon method has the highest reliability coefficient, namely 0.86; second, all reliable methods have a coefficient above 0.7 as the minimum requirement for acceptance; and third. All reliability methods have relatively the same reliability coefficient, namely around 0.8. fourth, all reliable techniques are suitable for measuring tools in the final semester exam. This research implies that lecturers are advised to use the odd-even Rulon reliability coefficient when measuring learning outcomes.