Linear weighting of sample survey data

Abstract

To improve the quality of estimates in sample surveys some kind of weighting is often carried out. Post-stratification is a popular weighting method. Two major problems can make applications of post-stratification difficult: empty strata and lack of adequate population information. This paper presents a general method for weighting, in which weights are obtained from a linear model which relates the target variables of the survey to auxiliary variables. Post-stratification is a special case of this method. Because of the generality of the method, different weighting schemes can be applied which take advantage of the available population as much as possible and at the same time avoid the mentioned problems. The theory is illustrated with an example.