The Research On Computerized Adaptive Test Based On Thegeneralized Partial Credit Model And 3 Parameters Luogistic Model

Effective integration of computer technology and psychological measurement,making computerized adaptive testing(CAT) is widely applied, traditional test(paper& pencil test)can not match the characteristics of high efficiency and flexible, etc.Currently, there are many researches on Graded Response Model(GRM), but few studies on Generalized Partial Credit Model(GPCM), GPCM has a huge advantage which requires step parameters are not monotonically increasing, so it conforms to the some of test situation.Selection strategy and termination rules are more intelligent in CAT. Selection strategy is the soul of CAT. The paper introduces a new method of harmonic average comprehensive selection, migrated good selection strategies under the GRM to the GPCM,,such as item selection with Exposure-Control factor, automatic control for the role of discrimination index and so on, tried to get whether the results are as good as CAT based on GRM or not.The paper studies how to quit of stratification in CAT which has few studies.According to a stratified based on 0-1 scoring 3PLM that raises a new termination rules, and it promoted to the b-blocked and a-stratified. Simulations show that the new termination rules better than the currently.