A note on the choice of sample size in estimating the mutual information

摘要

In a previous paper we have stated a criterion for choosing a suitable sample size to estimate the mutual information concerning two random variables in a finite population in random samplings. This criterion was based on Chebyshev's inequality. In the present paper we establish an alternative and conservative criterion based on the asymptotic distribution of the sample mutual information. We discuss the advantage in applying the new criterion, instead of the first one, in a practical example.