Sampling Survey Non-response Bias Adjustment

The question of non-response often appears in survey sampling, and is one of the difficult problem to deal with. Its emergence directly affects survey precision. Non-response is also a question that puzzles the investigation statistician. Although the foreign scholars of statistics have already do a great deal of research on the question of no-response, the current research on this issue in our country is still not attracted enough attention.Domestic research on this issue only of foreign literature and case study of sporadic, soithe research of non-response lack more systematic theory and effective methods to deal with. In the view of this, this paper from the adjustment of non-response bias, try to do the initial theoretical analysis and discussion on the question of non-response.First, the author systematically described the concept of non-response in investigate sampling, as well as the reason for the non-response, and from two point view of the deterministic view and stochastic view made a detailed analysis of the statistical impact on the estimator about non-response.Second, for the methods at home and abroad to deal with the item non-response and unit non-response, the author conducted a comprehensive analysis and systematic evaluation. there are some different quantity characteristics between response and non-response in actual survey, a new imputation method is proposed to make the imputation value close the real value, and adjust the imputation bias.At last, based on the analysis of the traditional double sampling method, a new method is presented to deal with the question of non-response, this method decompose the response layer, using stratification double sampling, for which do to improve the response rate . The optical solution is given under a certain cost of making the smallest variance.