Multivariate Generalizability Theory And Its Applied Research In The Design Of Faculty Evaluation

Generalizability Theory is one of the modern psychological and educational measurement theories. Multivariate Generalizability Theory is the deepening and development of the Univariate Generalizability Theory and is widely applied to all kinds of measurements. Based on systematically introducing the basic principles and technological methods, aimed at a multivariate nested unbalanced design of the faculty evaluation, this paper explores the methods of controlling errors through the Multivariate Generalizability analysis to the assessment of the variance-covariance components, and offers concerning practical reference value for the improvement of the faculty evaluation in different conditions. The study shows that the reliability(composite coefficient) of the total measurement is large, while the reliability(reliability-like coefficient) of some univariate measurements is sometimes not large, that students are out of accordance each other about the assessment of several teachers, that the priori weights are not the best, if they are changed we may get a higher Generalizability coefficient. The result also shows that thirty-five students at least are needed if we want to get a 0.8 composite Generalizability coefficient.