Mixed Logit Model And Its Application Research

Objective The purpose of this article is to research and compare the application of multinomial logit model and nested logit model and mixed logit model among discrete choice model in cooperative medical insurance model choosing behavior of rural residents in Zhejiang province, describe the similarities and differences of the three models in theory and application detailed. Aims to achieve the following objectives: mixed logit model will be introduced to the analysis and application of cooperative medical insurance mode selection; perform a comparison of multinomial logit model, nested logit model and mixed logit model, select an optimal model, analyze the affecting factors of cooperative medical insurance model choosing behavior of rural residents in Zhejiang province, and then provide a frame of reference to policy makers.Method The data considered were acquired from survey data of new rural cooperative medical in 2010 from Zhejiang province, and select the household heads as our study. Multinomial logit model and nested logit model and mixed logit model approach were applied to establish rural cooperative medical insurance mode choice behavior model, compare the differences of the goodness of fit index, the accuracy of predicted targets, the theory assumption of the models. The three models were established by using MDC procedure in SAS9.1. IIA hypothesis testings were completed with MLOGTEST module in STATA10.The simulation of the IIA assumption were done by Monte Carlo simulation techniques.Result Perform a IIA assumption testing based on multinomial logit model, the result can not satisfy the IIA hypothesis. The number of decision tree classification that can meet the utility maximization is only one. The Adjusted Estrella of the multinomial logit model and nested logit model and mixed logit model are 0.6430, 0.6427and 0.6658. The efficiency in the model predictions are 53%, 56%and 61%. Under both cases that IIA assumption were satisfied and unsatisfied, the alternative form of selection of rural cooperative medical insurance model choosing behavior established by multinomial logit model and mixed logit model were significantly different.Conclusion The mixed logit model is more appropriate than multinomial logit model and nested model to analyze the cooperative medical insurance modes choice behavior of rural residents in Zhejiang province. The factors affecting the decision of rural residents are family size, family labor force, education level, age, family income, the Engel coefficient, the number of...