不完全数据下响应变量的精度估计

Estimation Precision of Response with Incomplete Data

讨论了在不完全数据下的响应变量的估计精度,当不可忽略缺失下,样本的响应有缺失数据时,其联合分布是不可识别的.现证明了估计的精度与工具变量的相关性,当工具变量独立于目标变量时,目标变量估计的方差趋于无穷大.

The paper discusses estimation precision of an outcome with imcomplete data. When a sample of a binomial distribution has missing data from a nonignorable nonresponse mechanism, the distribution is not identifiable. The paper shows that the estimation precision is proportional to the instrumental variable,and the variance of estimation tends to infinite when the instrumental variable teads to be independent of the binary outcome.