3
$\begingroup$

Someone commented under my question Calculation of the moments using Hypergeometric distribution that $$ \sum_{k=0}^l\frac{{l \choose k}{2n-l \choose n-k}(2k-l)^q}{{2n\choose n}}\sim \sum_{k=0}^l (2k-l)^q {l \choose k}. $$

I've tried to use the Stirling's approximation formula, but I did not get the result. Maybe I have mistake... So is this approximation is true?

Thank you for your help.

$\endgroup$
  • $\begingroup$ Well, it doesn't seem to hold. What are the limits of $l, n$ that it is supposed to? $\endgroup$ – Eelvex Jun 12 '12 at 13:43
1
$\begingroup$

It's only true when $l = 0$ or $q$ is odd (since both sides are $0$), otherwise it is not true.

(I'll just do the case when $l$ is even, the case when $l$ is odd is pretty much identical.)

Since $\binom{a}{b}$ is at a maximum when $b = a/2$ (or $b = (a\pm 1)/2$ if $a$ is odd) and all the terms in the summation are non-negative ($q$ is even), we have

$$0 \le \sum_{k=0}^l \frac{\binom{2n-l}{n-k}}{\binom{2n}{n}} (2k-l)^q\binom{l}{k} \le \frac{\binom{2n-l}{n-l/2}}{\binom{2n}{n}} \sum_{k=0}^l (2k-l)^q\binom{l}{k}$$

$\binom{a}{a/2}$ is increasing, so it is maximised by taking $a$ as large as possible, that is, choosing $l$ as small as possible; so $l = 2$ ($0$ was eliminated above). Thus \begin{align*}\frac{\binom{2n-l}{n-l/2}}{\binom{2n}{n}} \sum_{k=0}^l (2k-l)^q\binom{l}{k} &\le \frac{\binom{2n-2}{n-1}}{\binom{2n}{n}} \sum_{k=0}^l (2k-l)^q\binom{l}{k} \\ &= \frac{n}{2n(2n-1)}\sum_{k=0}^l (2k-l)^q\binom{l}{k} \\ &= \frac{1}{2(2n-1)}\sum_{k=0}^l (2k-l)^q\binom{l}{k} \end{align*}

Thus, the approximation can't hold, since the left-hand side goes to $0$ as $n \to \infty$ while the right hand side doesn't.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.