Research On Latent Class Model And The Application In Missing Data Processing

Latent class analysis is an unsupervised statistical learning method that can be used to explain the complex associations between multiple manifest categorical variables,so that researchers can understand the potential influencing factors behind categorical variables through probabilities.It is often used in many fields such as sociology,biomedicine,and psychology.Missing data is a hot topic in related research fields in recent years,and the data collected in the survey often determines the quality of the analysis results.How to deal with the missing data problems encountered in latent class analysis is a hotspot and a difficult point in the analysis of latent class model.This paper first studies the latent class model,and combines the idea of multiple imputation method to deal with the missing data of the latent class model with dichotomous data.In the simulation experiment,we investigate the dataset under different missing rates(10%,30%and 50%)for the study of the impact of latent class analysis compared with the complete datasets.The result shows that the imputation method based on the latent class model performs better,and it is fit for the case of low missing rate.We apply the result of the simulation experiment to a psychological rehabilitation questionnaire survey on the drug rehabilitation personnel in the Yulu community of Kaiyuan City,Honghe Prefecture,Yunnan Province.After use the multiple imputation method for the questionnaire with missing values,the community can conduct targeted psychotherapy for the relevant population according to the best classification results by latent class analysis.The results of research show that the latent class model is valid for the problem of mental type division of drug abusers in Yunnan Province,which provides a reference for the new rehabilitation of drug rehabilitation in the rain dew community.