Estimation of the sample size required for obtaining given sample coverage

Abstract

We consider sampling from populations with large number of classes. The problem is to disclose a sufficiently big number of classes,
which represent a dominating part of population (e.g. 99%). In many applications, e.g. in genetics, disclosure of all classes is not necessary, since it can require a very large sample and, hence, is too costly. In this paper we propose a method for estimation of the sample size, necessary to achieve a given sample coverage. We apply the method to populations where the class probabilities are the members of a geometric sequence. A Monte Carlo study demonstrates that the method we propose gives good results for values if the common ratio of the sequence is not too close to 1.