Stratification of Univariate Populations: Does Distribution Matter?

The principal reason for stratification in the design of sample surveys is to reduce the variance of sample estimates. Efforts are always made to achieve further improvements in precision of these estimates. Not only does it depend on the sample sizes but also on the heterogeneity of the population units. Stratification in univariate populations has been explored by numerous researchers, many of whom have proposed competing algorithms that help surveyors determine efficient stratum boundaries. This paper takes into consideration the best criterion of using the estimated frequency distribution of the univariate population. To construct the strata boundaries, suitable cut-points are placed on the range of the estimated distribution using the dynamic programming algorithm which leads to globally optimum solutions. This paper investigates if the choice of the estimated distribution leads to different strata boundaries and hence contributes to the efficiency of the results. A numerical example using real data is presented to find answers to the question of whether the choice of distribution does or does not matter.