Sequential sampling for the comparison of domain means

Abstract

SUMMARY If it is required to compare the means of several domains of study in a finite population and if the domain to which each unit belongs cannot be ascertained before sampling, one may sample sequentially until specified precision requirements are met. Here, we determine the probability, P ( n ), that the precision requirements are satisfied by the time n observations have been selected randomly from the finite population. The dependence of P ( n ) on typically unknown design parameters is eliminated by assigning prior distributiona.