Research On The Models Of Discriminating Insufficient Effort Responding In The Situation Of Questionnaire

Insufficient Effort Responding(IER)not only can cause data pollution in the context of cognitive assessments,but also can pose a major threat to the quality of the questionnaire data,leading to significant deviations in the inferences drawn from the questionnaire data.Therefore,researchers pay more and more attention to the IER in the questionnaire.A series of methods aiming at detecting IER have been developed,but these methods still have shortcomings: Most methods only have good recognition ability for certain types of IER,but there are multiple types of IER simultaneously in the actual data set;Few individuals respond to items in a completely random pattern,current methods ignore state transitions within subjects,and fail to identify the subjects whose IER appear halfway;In the past,due to the limitations of questionnaire collection technology,Response Time(RT),as an important response process information,was only recorded in the form of total completion time,ignoring the role of RT in providing auxiliary information in identifying response categories during the response process.Therefore,this study draws on the methods of identifying abnormal behaviors in the context of cognitive assessments,and aims to provide a mixture model method that can simultaneously identify multiple types of IER and taking into account the process of the response state of subjects within the framework of hidden Markov models.Three studies have been carried out to this end: Study 1 provided a Hidden Markov IRT model based on RP data for paper-and-pencil response situation,and carried out simulation research to explore the model’s parameter estimation ability and IER recognition ability under different simulation conditions,and the model’s ability to identify subjects with different degrees of IER and different types of IER(including random IER,midpoint IER,and acquiescent IER)were discussed.The Hidden Markov IRT model based on RP and RT data was provided in Study 2 for online response situation,and the parameter estimation ability and IER recognition ability of the model were also discussed through simulation study.Study 3 is an empirical study,to illustrate the practical value of the model,using the proposed model to analyze the IER the IER in the data of the Critical Thinking Disposition Inventory,evaluating the reasonableness of the results,and providing specific operational means.The results show as follow:(1)The model proposed in study 1 can identify part of IER under the premise of ensuring the false positive rate.After adding RT data,the parameter estimation ability and IER recognition ability of the model in study 2 have been greatly improved.(2)The recognition ability of the model was most sensitive to the IER scale of the data and the RT distance between the two response states.(3)The recognition ability of the models decreased when the proportion of the subjects’ IER was very large or small.And the models have poor recognition ability for the acquiescent IER.(4)In empirical research,with the help of other information from empirical data,the rationality of the model analysis results was explained,and the occurrence law of IER was explored.And the calculation method of response effort index to divide different types of subjects and to enrich quality information of items was proposed,which provided practical reference for practitioners.