透過您的圖書館登入
IP:18.117.91.153
  • 期刊

伴隨變數有缺失下直線迴歸參數估計之有效性研究

Efficiency in Estimating Linear Regression Parameters with Missing Covariates

摘要


本文主要探討在一部份的伴隨變數有缺失下,使用完整資料分析法與迴歸校正法估計直線迴歸斜率參數時,有效性的比較,並提供迴歸校正法比完整資料分析法有效時,斜率參數應滿足的條件。本文也進一步引出加權的迴歸校正法,以使參數的估計最有效。我們用統計模擬說明在迴歸校正法比完整資料分析法有效的條件下,使用迴歸校正法估計斜率參數,其變異數確實比用完整資料分析法小,反之完整資料分析法則比迴歸校正法有效。從統計模擬的結果也可看出我們所提的加權迴歸校正法是三者中最有效的。對於參數估計量的標準差,我們則是利用拔靴法來估計,其結果的表現也與模擬的標準差很接近。

並列摘要


The purpose of this paper is to compare the relative efficiency of complete case analysis and the regression calibration method in estimating linear regression parameters when covariates are missing. We provide conditions under which the regression calibration method outperforms complete case analysis. Furthermore, we propose a weighted regression calibration method to improve upon the efficiency of either of the above alternatives. A simulation study investigates the performance of these three methods under different situations. Results show that the proposed weighted regression calibration method is the most efficient and that the standard errors estimated using a bootstrap procedure are satisfactory.

延伸閱讀