| The sample selection model was born in the research on women’s social problems,residents’ consumption,finance and insurance.In recent years,it has been used in the empirical research of various fields,such as economics,medicine,finance and so on.Tobit and Two-part model models are the most frequently used in the sample selection model,and they are also the econometric models most concerned by scholars.These two types of models are closely related,with both differences and connections.According to the author’s study and literature collection of the two types of models,it is found that most scholars’ research on these models focuses on empirical application rather than the analysis and innovation of model construction principles and estimation methods.Therefore,the purpose of this paper is to enrich the theoretical content and system of the sample censoring model by combing and summarizing the model construction principles and estimation methods of Tobit model and Two-part model.At the same time,through the discussion on the parameter estimation results and model selection methods under different circumstances,this paper provides the method and reference basis for scholars’ empirical research.This paper discusses the existing research results related to the topic,the construction principle related to the model and the statistical derivation process of the test method through the qualitative and quantitative research methods such as literature research method,Monte Carlo method,induction method and comparative analysis method,and then obtains the scientific and standardized experimental results through Monte Carlo simulation experiments in different situations,For analysis and summary.The innovation of this paper is to put forward the parameter estimation effect of Tobit model and Two-part model under different conditions,and put forward the discussion content of the "selection method" of the two types of model models,and carry out data simulation experiments for different contents and different programs.The conclusion of Monte Carlo simulation experiment in this paper is that when the sample deletion rate of other explanatory variables is larger than 50%,the original generation sequence is Two-part model,which is mistakenly set as Two-part model,the difference between the estimated value and the real value of the coefficient is large,which will affect the objectivity of empirical research and analysis;When the cut-off rate is small and less than 5%,the estimated value of the parameter is still close to the real value when the model is wrongly set.When the generated sequence is Tobit model,no matter what the estimated value of the censoring rate coefficient is very close to the real value,it does not affect the objectivity of the output result analysis in the empirical research.At the same time,Wald test statistics and LR test statistics can be considered to be able to effectively select the real generated sequence of the model at the significance level of 0.05,When the censoring rate is larger than 50%,it can be considered to be able to effectively select the real generated sequence of the model at the significance level of 0.01. |
