A super-population approach to multi-stage sampling

Abstract

This dissertation develops a new technique for the estimation of finite population parameters in a multi-stage sample survey. Specifically, estimators and confidence intervals for parameters of the finite population are developed for two-stage sampling when primaries are of either equal or unequal size, two-stage sampling when the variable of interest, y, is related to another variable, x, and p-stage sampling with an example of three-stage sampling when units are of equal size. The stochastic procedure generating the sample is assumed to be a two step procedure where the first step is selection of a "large sample" from an infinite super population and the second step is the actual implementation of the sample survey.