Nonresponse Bias Adjustment in Regression Analysis

DOI

この論文をさがす

説明

Nonresponse is an unavoidable problem in most sample surveys. If the proportion of nonrespondents is very small, nonresponse bias may be negligible. However, nonresponse rates in sample surveys have recently increased in many countries. Thus, methods for dealing with nonresponse bias are becoming an important topic. Regression analysis is often used to analyze survey data. In this paper, we discuss regression analysis with unit nonresponse. The least square estimator of regression coefficients may be asymptotically biased if nonresponse is not ignorable. In this paper, we establish a sufficient condition that a consistent estimator of regression coefficients is obtained. This condition can be determined from a causal diagram. Furthermore, we examine the results of this study by numerical experiments.

収録刊行物

詳細情報 詳細情報について

問題の指摘

ページトップへ