Determination of the Optimal Stratum Boundaries in the Monthly Retail Trade Survey in the Croatian Bureau of Statistics

Abstract

This paper aims to compare the current sample design of the Monthly Retail Trade Survey in the SAS environment and the sample design with generalized Lavallee-Hidiroglou method of strata construction in the R environment. The sample for the Monthly Retail Trade Survey consists of all enterprises with 10 or more employees and the stratified random sample of enterprises with less than 10 employees. Stratification is made by the principal activity and by the number of employees. Allocation is done by applying the Neyman allocation method and the allocation variable is retail trade turnover. Using the strata.LH function from the stratification-package in R, optimal stratum boundaries with restriction to a certain economic activity group are determined, using the number of employees as a stratification variable and the retail trade turnover as a survey variable. In one case, the same level of precision for the estimated total turnover is taken into account and in another one, optimal stratum boundaries with the given sample size are determined. Furthermore, the results with different models, which take into account the discrepancy between stratification and survey variables, are compared.

Keywords

• Monthly Retail Trade Survey
• strata construction
• optimal boundaries
• discrepancy