A Generalizability Coefficient Comparison of Thai Essay Test Scores for Phathomsuksa 3 Students Using Different Numbers of Raters, Scoring Methods and Scoring Patterns : A Case Study of Kamphaeng Phet Primary Education Service Area Office 2

Abstract

The purposes of this research were 1) to study the size of variance component compared with scores’ Generalizability Coefficient and 2) to compare deviation’s variance of Thai language essay test scores of grade 3 students with a different number of examiners, checking methods, and checking patterns. The research used the Multi-stage Sampling method to collect samples. The sample was 60 students in grade 3 of the academic year 2020 from schools under the Kamphaeng Phet Primary Educational Service Area Office 2. The tool of the research was a 6-item Thai essay writing test. Moreover, this research’s data was analyzed by Generalizability Coefficient with GENOVA. The results of the research concluded that 1. The variance in all conditions, test takers was the most valuable, followed by test and examiners respectively. 2. Scores’ Generalizability Coefficients with a different number of examiners, checking methods, and checking patterns were found that 2.1) Generalizability Coefficient with different numbers of examiners but same patterns and methods were found that the Generalizability Coefficient differed statistically significant at the 0.01 level except for checking some items of all test takers pattern. 2.2) Generalizability Coefficient with different checking methods but the same number of examiners and checking pattern differed statistically significant at the 0.01 level except for checking some items of all test takers pattern.                      2.3) Generalizability Coefficient with different checking patterns but the same number of examiners by discrete elements checking method was found that the Generalizability Coefficient differed statistically significant at the 0.01 level except for one examiner.  In part of Generalizability Coefficient with the overall assessment method differed statistically significant at the 0.01level except for 3 numbers of examiners. 3) The variance of test score’s deviation with a different number of examiners, checking methods, and patterns resulting that the deviation’s variance reduced when the number of examiners increased except for checking some items of all test takers pattern.