A Polytomous Extension Of Reparameterized Polytomous Attributes DINA

The DINA model has been applied widely in the field of education and psychometric.There are only two item parameters that affect the examinees’responses.Though there are lots of models have been developed to expend the DINA model,reparametrized polytomous attributes DINA model(RPa-DINA)can only deal with dichotomous data.An extension of RPa-DINA model was developed in study 1.The new model was named PRPa-DINA model.Whereas,Markov Chain Monte Carlo algorithms has been used to estimate the new model’s parameters.The independents of study 1 are sample size and item parameters(slipping and guessing).Karelitz(2004)and Peida Zhan(2016)has mentioned that polytomous attributes can translate to dichotomous attributes form.Then in Study 2,we were trying to prove the relationship of dichotomous attributes and polytomous attributes to simplify the process of parameter estimation.The independents of study 2 are simple size,the length of test,the highest level of attributes and the number of polytomous attributes.Results of simulation studies indicate that:(1)The MMR of PRPa-DINA model is higer than the PMR of it.(2)When the sample size is 500,increasing the item parameters are associated with classification accuracy and higher quality item parameters estimation.When the sample size is 1000,the lower level of item parameter is associated with classification accuracy and higer quality item parameters.(3)With the increasing of the level of item parameters while sample size is 1000,the accuracy of slipping and guessing are changed in two different directions.Guessing is tended to more accurate.(4)Polytomous attributes can treat as dichotomous attributes when the number of attributes is 3,but they are not equal when the number of attributes is 5.