With A Statistical Model Of The Missing Data Estimation And Testing

Missing data is prevalent in many studies, especially when the studies in-volve human beings. This phenomenon often leads to severe biases in the statisti-cal inference and analysis. How to deal with missing data, it has become a highly concerned problems. At present, it has a large number of documents appear at home and abroad. The paper has given related researches for specific statistical model and the statistical distribution under the condition of the missing data.This paper mainly is divided into three sections:The first part mainly introduces the missing-data mechanism and its re-search status, and introduces the semi-parametric regression model research sta-tus.The second part bases on the semi-parametric regression model with miss-ing at random under fixed design, the paper has given respectively the estimatorβ,gn(t),σn2 forβ, g(t) and error varianceσ2 and establishes their strong consis-tency under suitable conditions and qth-mean consistency using the least squares and the general nonparametric weighted method.The third part introduces two geometry populations with missing data, and discusses the maximum likelihood estimation. The strong of consistency and asymptotic normality properties of estimator are proved by Slutsky theory, and gives the population parameter of test statistics. Finally, its estimates of test and simulation are discussed.