Trends Stability of Reliability Coefficient Based on Sample Size and Ability of Test-takers

Abstract

One aspect that needs to be considered in the assessment of learning outcomes is the quality of the test by a stable reliability coefficient. This study aims to determine the trend of the stability of the reliability coefficient of the mathematics formative test based on the sample size and the ability of the test takers. The study was experimental in the form of a simulation, using a population of scores based on the answers of 403 test takers. The research sample was taken from the population of scores with 19 variations in sample sizes. Each sample size was repeated 31 times with the return technique; the reliability coefficient was calculated for each repetition and was used as the unit of analysis. In addition to the differences in sample sizes, the differences in the abilities of the test takers were also seen in two categories of high and low. Data were analyzed using exploratory-descriptive statistics and analysis of variance. Results showed as follows: first, the formative test of mathematics that was developed by the teacher at school has a reliability coefficient in the inadequate category; second, the reliability coefficient of the test tends to be more stable with increasing sample sizes; third, the difference in the ability of the test takers does not make a significant difference to the reliability coefficient; fourth, there is no interaction between sample sizes and abilities of the test takers on the reliability coefficient of the test.