Combining Information on Measurement Error in the Errors-in-Variables Model
研究了当解释变量存在测量误差时，工具变量法和衰减校正估计量的表现，发现测量误差不大时衰减校正法效果与强工具变量相当，且两者信息结合可得到更优估计量。
Abstract For the regression model in which an explanatory variable contains measurement error, the instrumental variable and the correction for attenuation estimators of slope are considered. Although each requires extra information that is rarely available in ideal form, examinations of asymptotic variances and simulation results suggest that the correction for attenuation performs as well as a fairly strong instrumental variable when the measurement errors are not too large, and this is often true even when the extra information acquired is rough. For larger measurement errors the correction method is of less value. Whenever both kinds of information are available, an estimator with smaller variance is possible by taking a weighted average of the two.