“…Let ;ij be the probability of inclusion of the jth unit of the ith stratum in the sample, j = 1, 2, ..., Nj; i = 1, 2, ..., k. As an estimator of the population total Y = EE Yij, consider the Horvitz-Thompson (1952) We call this the 0-optimum allocation. For this 0-optimum allocation we have a generalization of Rao's (1968) theorem the proof of which follows on the same lines as in Rao (1968) It is, however, not known from this under what conditions unstratified irPS sampling is still inferior to stratified irPS sampling when one deviates from the 0-optimum allocation. With this aim, Ramachandran and Rao (1974a) investigated whether stratified nPS sampling with various non-optimal allocations is likely to be worthwhile and whether it should at all be attempted in practice.…”