Estimation of Cronbach reliability based on sample size, gender, and the grades

Abstract

The study tries to see the effect of different sample sizes on the reliability of creativity tests using Cronbach’s alpha. In addition, the background of the sample (gender and grade) is also detected. The study has a practical significance for designing a sample for estimating reliability using Cronbach alpha. The secondary data are used for analysis. The sample of creativity test is determined purposively, that is, has been administered to 1,960 students. The α-coefficient calculation uses the SPSS program. The result proves in the total sample, the creativity test is reliable (α=0.852). In small sample sizes, the α-coefficients values vary greatly in the different test groups. In the larger sample size, the α-coefficients are closer to constant in the different test groups. For a polytomous scale test with a length of 20 items, the α-coefficients obtained close to constant on the sample size of more than 80 testees. This result indicates for the optimum result the sample size is at least four times larger than the number of test items in reliability analysis, especially Cronbach alpha. Based on the gender aspect, descriptively the α-coefficient in the male group (0.846) was almost equal to the female group (0.851). By grouping the testees based on the grades, there are slight differences in α-coefficients in class IV (0.850), class V (0.841), and class VI (0.809). Regarding the results of this study, in the reliability estimation of test development, it is suggested to consider the adequacy and the characteristic of samples.