Many researches on determination of stratum sample sizes have been made and final sample allocation is conducted using a variety of methods. When a population is divided into several strata, the sample allocation that minimizes the variance of the whole population is Neyman allocation. However, in most cases, Neyman allocation is not an optimal sample allocation since most of the sample surveys require that each stratum size should meet a certain level of relative standard error. Many sampling designs for official statistics use various methods, such as power allocation and minimum sample allocation, to determine the sample size that satisfies a given stratum criterion. In this study, we suggested a sample allocation method that can reduce the relative standard error of the whole population while satisfying a given relative standard error criterion of a very small size stratum. Also We conducted small simulation studies to explain the characteristic of the suggested method.

Key words : Neyman allocation, Power allocation, Coefficient of variation, Relative standard error