Generalizability Theory and Comparison of the Results of G and D Studies Computed by SPSS and GENOVA Packet Programs

Abstract

In generalizability theory, it is possible to estimate one reliability value according to several different sources of measurement error and their interaction into one study. However, in classical test theory, different reliability values can be obtained for situations in more than one source of variance. Generalizability theory is an extension of the classical test theory which considers multiple sources of measurement error simultaneously. Analyses related to generalizability theory were generally done with GENOVA computer packet program. Because using the program is hard and complex to understand, it is a major problem in studies related to generalizability theory. Musquash and O'Connor (2006) described to use SPSS (also SAS and MATLAB) for conducting the analyses about generalizability theory. In this study, generalizability and dependability coefficients for both generalizability (G) study and decision (D) studies were presented by using both GENOVA and SPSS computer packet programs. This study also provides an illustrative example and comparison of the results on both programs.