The Application of Generalizability Theory In Constructing Achievement Tests

Abstract

The Application of Generalizability theory to test construction provides a flexible and practical framework and yet still remains inaccessible to test developers because of its technical and mathematical approach to measurement. Different methods of test construction to measure mastery of a universe in cognitive domain have been proposed. Simple random and stratified t random sampling procedures are two such models. The focus of this study was to examine the application of Generalizability Theory when tests have been constructed using the two models and administered to a group of students. Both coefficients alpha (6) and alphas (ft) as indices of generalizability were computed. The findings indicated that a test constructed using simple random sampling procedure had a better coefficient of generalizability although within the comparison analysis indicated that taking stratification into account increased the generalizability of the test and that generalizability theory could be" applied to the construction of achievement tests.