对于0-1评分模型,R.B.Juan等提出了最大信息量分层选题策略,将此选题策略应用到等级反应评分模型(GRM)中,即以项目j的最大信息量I max(j)作为分层的依据,以取得该项目的最大信息量时能力点值θmax(j)作为项目的综合难度,分别用I max(j)与θmax(j)替代张华华等提出的按a分层和按b分块按a分层方法中的a、b参数,形成最大信息量按a分层选题策略(MI-AS)和最大信息量按b分块按a分层方法(MI-BS).模拟实验结果表明:MI-AS和MI-BS方法较传统的按a分层方法要好.
The stratified item selection strategy with maximal information proposed by Juan R.B.is applied to graded response model.Item j’s maximal information I max(j) is taken as the basis of stratificaton,and the ability value θ max(j) where the item j reaches its maximal information is taken as item j’s comprehensive difficulty.The parameter a and b in the a-stratified method(AS) and a-stratified method with b blocking(BS) proposed by Chang Huahua et al.are replaced by I max(j) and θ max(j),respectively,so as to form MI-AS method and MI-BS method.The simulation experiments demonstrate that MI-AS and MI-BS are better than traditional AS and BS.