Choosing a parameter for a Poisson distribution

49 Views Asked by At

This question is kind of weird, and it raised from an even weirder question (in the bins and balls model), but I've tried to simplify it as much as I can

let $n,k\in \mathbb N$ be 2 numbers such that $\log n\le k\le\sqrt n$, our job is to choose a number $x\in \mathbb N$

such that if $X\sim \text{Poisson}(x)$ we have

$\binom {2n}{2x}\Pr [X\ge k]\ge 1$

But! we want $x$ to be as low as possible

for $x=2k$ we have (by Markov's inequality) that probability is at least $\frac{1}{2}$ and the Binom coefficient is some positive integer $\ge 2$, so the inequality is satisfied, but I hoped that one could achieve something better, more specifically $x$ that is asymptotically less than $k$

It seems one cannot use the Chernoff bounds here, because we want to lower bound the probability of $X$ to be more than its expectancy