Determination of Type I Error Rates and Power of Answer Copying Indices under Various Conditions

Author/s: Seha Yormaz, Önder Sünbül

DOI: 10.12738/estp.2017.1.0105  OnlineFirst published on November 15, 2016

Abstract

This study aims to determine the Type I error rates and power of S1, S2 indices and kappa statistic at detecting copying on multiple-choice tests under various conditions. It also aims to determine how copying groups are created in order to calculate how kappa statistics affect Type I error rates and power. In this study, option-matrices were obtained by using five choices of multiple choice test data generation. R-3.0 software was used for data generation and analysis. The effects of sample size, test length, and source ability level on Type I error rates and also amount of copying on power according to nominal α-levels were investigated. The findings indicated that Type I error rates for S1 and S2 indices were lower in all conditions, while the same error rates were relatively higher for copy detection using kappa statistics. Kappa statistics were seen to be more powerful than S1 and S2 indices under all conditions. As sample size, test length, and amount of copying increased, the S2 index was also revealed to be more sensitive than the S1 index. Type I error rates and power results of the kappa statistics were similar to each other.

Keywords
Copy detection methods, S1 and S2 indices, Kappa statistic

Click for Turkish (Türkçe)