在统计分析中，数据缺失的现象是非常普遍的。变量选择是统计数据分析必不可少的工具，对缺失数据和变量选择的研究已有许多研究成果。对于缺失数据问题的变量选择问题是新的统计挑战问题。近几年，已有针对此问题的一些研究成果，已得到的研究成果中，多采用SCAD和自适应LASSO方法进行变量选择。. 本项目将用Bayes方法研究缺失数据的变量选择，主要研究以下的问题：(1) 用Bayes方法对线性回归模型的协变量、响应变量在随机缺失的情况进行变量选择；(2) 对随机效应模型，研究协变量不同缺失机制的变量选择问题，在确定先验分布后，对所有参数的后验分布进行推导。(3) 用Bayes方法对具有缺失数据的对数线性Poisson模型进行研究。(4) 通过大量模拟计算与其他变量选择方法结果进行比较，说明项目给出方法的可行性和优越性。(5)进行实证分析。

In statistical analysis, missing data is a very common phenomenon. Variable selection is an essential tool for statistical data analysis. A large amount of theoretical and applied studies on the missing data problem and variable selection have been present in the literature. Performing variable selection in statistical models for missing data problems raises several new statistical challenges. Recently, some work has been done on variable selection with missing data. I In the obtained research results, SCAD and adaptive LASSO method have been applied to different models for variable selection.. In this project, we will consider follow problems. We will study the linear regression　models when the response variable or covariate is sometimes missing and study the　random effects model with covariate missing. The missingness structure is missing　at random. For a prior determination distribution, the posterior distributions of　all parameters are derived. We will apply Bayes method to study the Log-linear Poisson model with missing data. In this project, simulation studies and real cased are considered.