مرکز منطقه ای اطلاع رساني علوم و فناوري - Examining Split-Half Reliability for Sample Sizes In Different Homogeneity Levels According to Test Length, Split-Half Methods And Reliability Estimation Techniques

Abstract :

The purpose of the study is to examine split-half reliabilities of tests in different lengths for low, medium and high homogeneity level groups based on different split-half methods and reliability estimation techniques. The unidimensional data simulation of the study was generated on Wingen 3 software depending on 2 Parametric Logistic Model of Item Response Theory. In order to determine how split-half reliabilities is affected by sample size, 50 samples for 50, 300 and 500 sample sizes were drawn from populations by using simple random sampling. The split-half reliabilities and the standard errors of sample reliabilites were computed by codes written in MATLAB 6.5 programming language. On examining the results, it was found that the more heteregoneous is the sample, the more higher the sample and test length, the higher is the split-half reliabilities and the standard errors of reliabilities were decreased. When the tests split into halves with odds-evens method and odds-evens according to item difficulties method, the split-half reliabilities were higher; the standard errors were lower. The difference between the standard errors and the medians of split-half reliabilites attained by reliability estimation techniques in all homogeneity groups was scarcely any. It was found that the consistency between populations and samples was higher in the cases where the groups were heteregenous; the test length and the sample sizes were high and odds-evens method and odds-evens according to item difficulties method were used.