基于DA插补法的线性回归模型系数估计值的模拟研究

Data Augmentation(DA)插补法是最常用的MCMC多重插补法之一。利用模拟方法研究基于DA插补法的线性回归模型的系数估计值,分析估计值的统计性质受无回答机制、无回答率和插补重数的影响。模拟结果显示:在完全随机无回答机制下,选择较小插补重数常常会得到较好的回归系数估计值;在随机无回答机制下,随着无回答率增大而选择更大插补重数往往会得到更好的回归系数估计值;在非随机无回答机制下,选择更大插补重数并不一定总会得到更好的回归系数估计值。

On Estimators of Coefficients of Linear Regression Models Based on Data Augmentation Multiple Imputation

Data Augmentation （DA） multiple imputation is the most commonly used version of the MCMC multiple imputation. This paper simulates coefficients estimators of linear regression model based on the DA multiple imputation and analyzes their statistical properties when non-response mechanism, non-response rate, and number of imputation are given. Simulation results show, it is usually the better coefficients estimators with the small number of imputation at Non-response completely at random mechanism. At non-response at random mechanism, as the non-response rate becomes larger, it is often the better coefficients estimators with the greater number of imputation. At not non-response at random mechanism, the better coefficients estimators is not always follows greater number of imputation.