Journal of Official Statistics, Vol.12, No.3, 1996. pp. 241–251
Abstract:From a finite, two-dimensional population a sample of rows and a sample of columns are drawn independently. A sample from the two-dimensional population is formed by cross-classifying these samples. We call this procedure cross-classified sampling (CCS). This article is concerned with the problem of calculating the variance of estimators based on CCS samples. A general variance decomposition is presented. This decomposition leads to our main result, which simplifies the variance derivation when the one-dimensional samples are (separately) stratified. Finally, we apply the general results to the special case with the Horvitz-Thompson estimator in unequal probability sampling. The results in this article have been used to derive variance estimators for the Swedish Consumer Price Index.
Keywords:Two-dimensional sampling; variance decomposition; variance estimation; Consumer Price Index.
Copyright © Statistics Sweden 1996-2018. Open AccessISSN 0282-423XCreated and Maintained by OKS Group