| Partially linear model has been developed an important semi-parametric model instatistics. It is commonly applied into economic statistics, medical statistics and analysis offinancial time series. Meanwhile, the model is effectively able to avoid “dimension disaster”problem in general semi-parametric regression model.Missing data exists in several region. It is universally researched by specialists. As aresult, the problem is significant to study.In current literature, many scholars have got achievement of the estimate methods andthe asymptotic properties of partially linear model with missing response. Nevertheless,region of testing serial correlation of the model is insufficiency. We usually demand thefitted residuals are white noise for a good fitted model, meanwhile the residuals satisfy theterm of independent and homoscedasticity. If error serial of modelis serial correlation, theymay cause many troubles, as estimator is invalid, model is inefficiency etc. The thesisconsider the issues, testing for serial correlation in partially linear model with missingresponse is shown. The framework of thesis as follows:The first is introduction. We show missing data, testing correlation, empiricallikelihood and current achievement of semi-parametric model with missing data.The second part consider the estimate method of partially linear model with missingresponse and asymptotic properties. The part mainly pave the way for proof part.The third chapter is the major. In this section, we use the method of VT,ptest andempirical ratio test in testing semi-parametric model under missing response. Meanwhile,we got the asymptotic distributions under null hypothesis and assumptions. Ultimately,results of simulations illustrate both tests own good properties with finite sample.The fourth has proved the theorems and conclusions of third part.The fifth has applied partially linear model in study of wheat yield.The sixth has made summary and outlook of semi-parametric model with missingdata. |
PDF Full Text Request