Question 1: Are the following empirically observed relationships true

$$ {n \choose 1^a}{n \choose 2^a}{n \choose 3^a}\cdots {n \choose m^a} \sim \exp\bigg(\frac{2n^{1 + \frac{1}{a}}}{a+3}\bigg) $$ where $a$ is a fixed positive integer and $m = \lfloor n^{1/a}\rfloor$.

$$ {n \choose b}{n \choose 2b}{n \choose 3b}\cdots {n \choose mb} \sim \exp\bigg(\frac{n^{2}}{2b}\bigg) $$ where $b$ is a fixed positive integer and $m = \lfloor n/b\rfloor$.

Question 2: What is the growth rate of

$$ {n \choose 1^ab}{n \choose 2^ab}{n \choose 3^ab}\cdots {n \choose m^ab} $$

For $a = 3$, the $\%$ error between the asymptotic and the actual product is shown below. We observe that the error is small and is decreasing with $n$?

Note: Posted in MO since it was unanswered in MSE.

enter image description here

  • 1
    $\begingroup$ what does Stirling's approximation say here? (I understand there are too many factorials) $\endgroup$ – Venkataramana Aug 12 '19 at 3:25
  • $\begingroup$ in the second one you probably have $k$ and $m$ confused $\endgroup$ – Brendan McKay Aug 12 '19 at 7:20
  • $\begingroup$ @BrendanMcKay corrected $\endgroup$ – Nilotpal Kanti Sinha Aug 12 '19 at 7:53

Partial answer (but the general case should be similar): using Stirling+Euler-MacLaurin+Glaisher's constant we have for $a=1$: $$\prod_{1\le j\le n}\binom{n}{j}\sim n^{-(n/2+1/3)}e^{n^2/2+n(1-\log(2\pi)/2)+K}$$ with $$K=1/12-2\zeta'(-1)-\log(2\pi)/2\;.$$ Thus I presume that $A\sim B$ in the OP's question means that $\log(A)/\log(B)$ tends to 1.


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.