Investigation of Measurement Precision and Test Lengths in Computerized Adaptive Tests in Different Conditions

Abstract

In this study, it is aimed to examine item exposure rate, content balancing, and ability estimation in terms of termination rules with regard to testing lengths and testing accuracy in computerized adaptive testing. In this context, EAP and MLE ability estimation methods were compared with 1, 2 and 4 group content balancing pattern; 0.50, 0.75 and 1.00 exposure rate; it was compared with a total of 72 different conditions, including 0.35 and 0.40 standard error-based and the termination rule based on the test length of 15 and 30, was compared to correlation, bias, RMSE and test length. The production and analysis of the data were performed in the R program. As a result, the best performance in the measurement is a fixed test length of 30 items with 0.35 standard error; in group 1 pattern where the content balancing is not a group limitation; the exposure rate was displayed in the range of 0.75 and 1. Depending on the test length of ability estimation methods, scope balancing patterns and exposure rates, the number of items changes in the range of 22 and 25; Based on the termination rule, it was estimated that at least 0.40 standard errors with a standard error based on 39 items.