Research Of Internet Addiction In Computer Adaptation Test Based On Multidimensianal IRT

Internet addiction(IA)is commonly refers to the behavior of dependence and adverse consequences due to inappropriate use of the network.The symptoms can be expressed as follows: losing interests in reality;Exceeding the general limit of online time to achieve the psychological satisfaction;Failing to control themselves regardless of the severity of the problem.The traditional scale for IA is based on the Classical test theory(CTT),which are unavailable to implement testes in a personalized way on one hand,and can't achieve an objective and accurate measurement of the participant or the item itself on the other hand.But the Item response theory(IRT)and computerized adaptive test(CAT)can fundamentally overcome the disadvantages of CTT mentioned above.Based on the(Multidimensional item response theory)MIRT and CAT,This study will develop a new Internet addiction measurement,In one aspect,building the self-adaptive measurement of different levels of addict so that the results can be more accurate;in the other aspect,reducing the test length without loss of test accuracy,thereby the test burden will be alleviated.This study includes three sub-studies:Study 1: Under the frame of bifactor-model,the dimension structure of Internet addiction is analyzed by EFA and CFA method.Study 2: Under the frame of multidimensional item response theory(MIRT),select high quality items according to item-fitting index,degree of differentiation,DIF and load factors loading index to build the IA bank,so that we can lay the foundation for study three.Study 3: With MIRT and CAT,MCAT-IA and its related algorithms are develop-d and the relevant algorithms will be verified.Finally,the reliability and validity of MCAT-IA will be further tested.The results show that:(1)in the EFA and CFA analysis,bifactor-model with three-dimensional structure(a common factor and three special factor)fit best in all compared models.(2)123 items in the item bank have a factor loading above 0.3 and differentiation parameter greater than 1,the average degree of differentiation is 1.567(SD = 0.265).the result indicates that all items have a good psychometric properties.(3)using Monte Carlo simulation to simulate the response in the MCAT-IA,and adopted the Bayes D optimization selection strategy and EAP method to estimate the ability parameter of participants,the results on G factor show a high correlation between the results from MCAT and using of all items in the bank(0.905 when use 20 items).(4)using Fisher information and standard error to test the reliability of MCAT-IA,which proved that the MCAT-IA can reach the result of 123 items using 25 items with-out decrease the accuracy.It saved as much as 79.7% of the test length,so it can undoubtedly reduce the test burden.(5)with the increase of used number of items,the correlation between MCAT-IA results and the calibration scale IAT results,as well as the self-reported of IA degree and the usage time is increasing;At the same time,all correlation coefficients were between [0.3,0.8],reaching the medium and above correlation in psychometrics(p<0.001).(6)finally,this study was based on bifactor multidimensional model developed MCAT � IA,which has an ideal validity and reliability.So it provides a new measurement technique and tools for practical application.