Answers are grounded in this paper and cite their sources.
We compare multivariate hypergeometric sampling with multinomial sampling from the same urn. For a population of size n, a sample of size k, and positive color counts l_1,...,l_c, the relative entropy equals the sum of the binomial-to-Poisson relative entropies b_(l_i)(k/n), minus G(k/n), up to an explicit O_delta(c/sqrt(n)) error, uniformly over all color compositions for delta <= k/n <= 1-delta. Here G(t)=(-log(1-t)-t)/2. An exact identity for a binomially randomized sample size and a composition-uniform increment bound yield the comparison. The formula retains fixed, arbitrarily slowly growing, and macroscopic color counts. A finite singleton-color counterexample addresses a previously stated uniform lower bound.
No takes yet. Share an insight, caveat, or question.
Zihuan Xu (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: