Abstract
In our study, we examined the reliability of students' evaluation of university teaching. First, we analyzed four-facet data (teacher×rater×viewpoint×order) by the generalizability theory approach, and estimated the variance components for the facets as a G study. Then, we evaluated the reliability from the point of view of the generalizability coefficient (a reliability-like coefficient that is used when a decision concerns the relative ordering of individuals) and the index of dependability (an index that is used when a decision focuses on the absolute level of an individual's performance independent of others' performance), and examined how those indices change as the number of raters and viewpoints are changed. With these analyses we found conditions needed to maintain sufficient reliability of the evaluation for various situations. We also introduced a new method to evaluate the reliability at each level of facet by using structural equation modeling. By this method, one can examine the reliability of students' evaluation of university teaching more specifically.
