The Analyse Of Oral Rating-time With Generalizability Theory

This dissertation includes three main content.One is to introduce Generalizability Theory to evaluate the reliability of subjective rating.GT has a unique role to evaluate the error of measurement.It is based on variance component analysis and applies the technique of statistical adjustment to evaluate the errors that come from the different side of rating in one measurement.Then it can provide two reliability coefficients which can be used in different decision tasks,one is generalizability coefficient, Eρ~2 and the other is dependent Index, Φ .The second one is to introduce therating method of oral .Third one is to apply generalizability theory to analyse oral rating-time.Six raters score sixty HSK(advanced) oral with a twelve-point scale.Then we analyse the data which we obtained.In this part we have two step,G study and D study.In G study we found examinee interaction had high proportion,so we can say the rating reflects oral competence of examinees in deed.In D study we obtained two high generalizability coefficients of T1 and T2.They can be accepted.But T2 has higher generalizability coefficient.