The Development And Application Of Nonparametric Cognitive Diagnosis Models For Polytomous Data And Use It In CD-CAT

Cognitively Diagnostic Assessments(CDA)combine the advantage of cognitive psychology and educational measurement which are designed to provide related diagnostic information to teachers and students to modify teacher teaching strategies.The existing polytomous parametric cognitive diagnosis models(CDMs),such as the sequence processing G-DINA model(seq-GDINA;Ma & Torre,2016;BJMSP),work well for large dataset,however,their parameter estimation accuracy cannot be guaranteed in small-scale educational and psychological measurement due to that they need large sample size to guarantee reliable model parameter estimation when using the statistic algorithms(such as MMLE-EM or MCMC algorithm).In addition,except the seq-GDINA,none of these polytomous CDMs consider the relationship between attributes and response categories.To address this issue,this paper proposes a sequential general polytomous nonparametric classification method for small-scale measurement.In order to take full advantage of the efficiency,speed and accuracy of Cognitive Diagnosis Computerized Adaptive Testing(CD-CAT),this study is also discussed in a small sample,Nonparametric CD-CAT for polytomous data algorithm.In view of this,this article carries out three studies:Study1:Nonparametric cognitive diagnosis models for polytomous data is proposed.The proposed method remedies the assumption of constraint model with weighted ideal category response and take category Q matrix into method.Our simulation results suggested that under small sample size context the proposed nonparametric method outperformed the existing general polytomous parametric CDM(i.e.,seq-GDINA model)in terms of the classification accuracy.The higher the quality of the question,the higher PAR of seq-GNPC model.In addition,we carried out the nonparametric cognitive diagnosis models for polytomous data in empirical data.The real data analysis is also demonstrated the proposed model provides more reliable classification.Study 2: Nonparametric cognitive diagnostic computerized adaptive test for polytomous data is proposed.First,the dichotomous items selection method and the nonparametric item selection with constraint model extended to the seq-NPCD-CAT,namely,seq-GNPS,SEQ-HDWIR and seq-DWIR.The simulation is focus on the effectiveness and efficiency of these compared methods in classifying examinees and their ability to control item exposure with small calibration samples.The results of seqNPCD-CAT show that: First,the test security of seq-GNPS is the highest.Secondly,the PAR estimated by the three selection strategies was close to 1 when the calibration samples were small and item length at 20.Study 3: We carried out the nonparametric cognitive diagnostic computerized adaptive test in empirical data.The results showed that the three nonparametric polytomous data item selection method have higher test security.In addition,only 20 items can achieve the estimation accuracy of the symptoms of internet addiction.