Konstrukcja estymatorów kalibracyjnych wartości globalnej dla różnych funkcji odległości
Construction of calibration estimators of totals for different distance measures

Summary/Abstract: Missing data are one of the major types of non-random errors in statistical surveys. One of the methods proposed by Deville and Särndal [1992] which is designed to offset the negative effect of missing data is calibration, which is successfully used in practice by statistical offices of many countries. In its classical form calibration is a method in which calibrated weights are computed by minimizing a distance measure between the initial sampling weights and new weights, which need to satisfy certain calibration constraints. The main goal of this paper is to present the construction of calibration estimators of totals for different types of distance measures. Its empirical part, based on the calib function, which is available in R program in the sampling package, is devoted to the method of finding calibration weights in surveys with nonresponse for different distance measures.

• Details
• Contents