The Comparison of Reliability According to Generalizability Theory and Classical Test Theory on Random Data

Abstract

In this study, a framework was tried to be presented by focusing on the similarities and differences of generalizability theory (GT) and classical test theory (CTT). Although this example consists of 125 students, 18 items, and 4 raters, all data was obtained completely in a random way. This completely random data which was different than any data previously used in the research available was used to assess and compare the GT and CTT. The results reflects that the Cronbach's alpha and G-coefficients are very low and the same for a single facet design (s x i) and Phi-coefficient and (1) coefficient were obtained as .457 and.456, respectively for two facet design (s x i x r) which was not examined by CTT.