On Matrix Sampling And Comparative Study Of Ability Parameter Estimation

In most large-scale educational achievement several broad content domains are tested.Since it is impossible for students to accomplish the all content domains,the large-scale educational use the balanced incomplete block design.A popular design is the balanced incomplete block design(BIB).While PISA?NEAP?TIMISS use the balanced incomplete block design,Massachusetts and Kentucky use-equivalent groups with anchor test design.This study evaluates the accuracy and stability of the ability estimation between traditional estimation methods such as EAP and the plausible value estimation method under the balanced incomplete block design.The simulation experiment consisted of two studies:(1)Comparison of the mean and standard deviation estimation accuracy and stability of the traditional methods and the PV method when the background characteristics of students was unknown;(2)When the background characteristics of student was under the circumstances of knowledge,the comparison of traditional methods such as EAP and plausible value estimation method on population mean and standard deviation estimation of accuracy and stability.The first experimental situations show:(1)the EAP method and PV method better than that of other methods(MLE?WLE)in balanced incomplete block design,and as test length increase,population parameters(means and standard deviations)are well estimated;(2)The accuracy and stability of the EAP,MLE,and WLE methods for estimating the population mean and standard deviation will increase and increase with the total number of students,but the accuracy and stability of the PV will increase in a relatively small range and the ABS and RMSD tend to reach a certain level.The second experimental situations show:(1)the PV method better than that of traditional methods(MLE?WLE)in balanced incomplete block design even in the condition of small size and short test length.(2)there is a great difference in the accuracy and stability of the group estimation between the PV estimation method is added with background information and without background information.