抽样调查中计量误差的分解及应用

Decomposition and Application of Measurement Errors in Survey Sampling

建立了以计量误差作为随机变量的一般性计量误差模型,从方差和计量偏差两类误差效应来对误差进行分解,进一步从方差中分离出计量方差,包括简单计量方差和相关计量方差,以及抽样方差,并得出上述误差的计算公式和估计方法.针对抽样实践中最常见的调查员误差,结合调查员分配方式,提出使得重复调查中计量结果的相关系数尽可能小是控制相关计量方差的关键,而增加子样本中调查员的数量可以有效减少调查员方差.

With measurement error modeled as the random variables, a general measurement error model is presented. It decomposes the errors into variance and measurement bias. The variance can be further decomposed into measurement variance and sampling variance. The measurement variance is composed with simple measurement variance and correlated measurement variance. The calculation formula and estimation method of the above errors are obtained. According to the most common interviewer effects in survey practice, combined with investigator assignments, it is proposed that the low correlation coefficient of measurement results in repeated surveys is the key to control the correlation measurement variance, and that increasing the sub-sample interviewers can effectively reduce investigator variance.