Research Of Random Missing Data Treating Methods In Psychological Questionnaire

Questionnaire is one of the most important research methods in the field of psychology and education,that it can collect a lot of research data in a relatively short time makes it deeply loved by masses of psychology and education researchers.When we use questionnaire to collect data,there are missing data in the dataset,because of the subjects' unwillingness,time limit and other subjective or objective reasons.a small amount of missing data may not affect the accuracy of data analysis and research results,when the amount of missing data is large,and it is of difficulty to take remedial measures to recollect the missing data,it is wise to take the mass missing data into consideration,At this time,the missing data need to be handled appropriately so that accurate parameter estimates and statistical inference based on population parameters can be obtained.In previous studies,researchers have mostly adopts the inferior missing data treatling method:incorrect(IN)or list deletion method(LD)to analyze the data of the questionnaire which consists of missing data,but studies have shown that the effect of these methods are not ideal.At the same time,in learning tests or ability tests such cognitive tests,EM and MI is very effective in dealing with missing data.So this paper probed the effectiveness of these methods in the questionnaire data analysis.Above all,the current study aims to address this question by exploring several missing data treating methods' effectiveness under different proportions of missing data(1%,5%,10%,20%,30%)and number of categories in the items under the two methods of questionnaire data analysis.The research includes two simulation studies and an empirical study,implications of these findings and future research directions are discussed.The first study is to explore proper missing data treating methods in the questionnaire data analysis under classical test theory;the second simulation study is to explore the proper missing data processing method in the questionnaire data analysis under item response theory;the third study is an empirical research and to test the accuracy of the results from the two simulation studies in practical application.The main results are as follows:under the classical test theory method,the degree that subjects ability parameters affected by the independent variable is greater than the item difficulty and discrimination parameters,as the missing rate increases,the parameters of the subjects' ability and the questionnaire's item difficult and discrimination parameters is becoming more biased,when IN is used to deal with the missing data,the BIASabs increase with the categories,that is,the more categories,the bigger the BIASabs value;LD can produce unbiased ability parameter,but subjects who have missing data in their response vector will not have ability parameter in the condition of LD treatment,and MI have a good effect in dealing with missing data,at last,EM is also a well alternative method.Under the itm response theory,the discrimination parameter is more susceptible to the dependent variable than other parameters,especially under high missing rate(20%or more);besides,The higher the missing rate,the greater the BIASabs value;Under the IN method,the BIASabs increases with the categories in items;MI method were better than others under various conditions.