Empirical Likelihood Estimation Of Conditional Quantile Under Missing Data

Statistics is a science about data, the existing theories and methods are mostly built on the basis of statistical analysis which is based on the complete samples,However, the missing data situation is prevalent in the real fields, In the fields of market research, opinion polls, survival analysis, medical research, the condition of missing data is often appear, The main reasons are that some sample individual is unwilling to provide the required information, the data of loss due to some uncontrollable reasons., Researchers’mistakes maybe also cause missing data, etc, In these cases, general statistical methods can not be directly applied to incomplete statistics, Therefore, there has important practical significance to study the statistical properties of some important parameters such as conditional quantile in the case of missing data.In this dissertation, the work is done by the following:Firstly,we study the estimation of conditional quintile when the response variable meet missing at random mechanism via the empirical likelihood method, and the conditional quantile of the asymptotic normal distribution in the regularity condition, Secondly, In response to the variables in the random missing mechanism, the confidence interval of conditional quantile in the absence and presence of some anxiliary information is given on the basis of empirical likelihood. It is also shown that the asymptotic efficacy of test is not be reduced as the information is added, Finally, the simulation of the theoretical results has be done by the use of mathematical software.This thesis is supplementary to the existing literature and the promotion of the previous results.