Statistical Inference from Complex Sample with SAS on the Example of Household Budget Surveys

Abstract

Many sample surveys are not based on simple or unrestricted random samples, but usually on complex samples with stratification, clustering, unequal inclusion probabilities and multistage sampling. To estimate a parameter, all individual data from complex sample must be weighted by weights connected with sample selection scheme and statification and adjusted for nonresponse and noncoverage errors. Standard statistical computer software is correct for statistical inference from unrestricted random sample, but not from complex sample. The aim of this paper is to present utility of SAS software to statistical inference from complex sample. Data from Household Budget Survey 2008 were used in examples.