6
$\begingroup$

From the Berry-Essen theorem I can deduce

$$\sup_{x\in\mathbb R}\left|P\left(\frac{B(p,n)-np}{\sqrt{npq}} \le x\right) - \Phi(x)\right| \le \frac{C(p^2+q^2)}{\sqrt{npq}}$$

with $C \le 0.4748$.

My Question: Is there a better estimate for the constant $C$ than the one given above for the special case of the binomial distribution?

Reason for my question: The given inequality for $C$ holds for any standardized sum of any i.i.d random variables. But I am only interested in the case of binomial distributed random variables. From the answer to my question Estimates for the normal approximation of the binomial distribution I know, that I cannot except a better estimation. But I guess, that there is a better estimate of $C$ if one restricts the Berry-Esseen theorem to binomial distributions only. It would be great when you can point me to an article or a textbook with a better estimate of $C$.

Note: I am aware, that there are better approximations for the binomial distribution than the normal one. See this answer.

I already asked this question on stats.stackexchange.com, but I didn't get an answer there. I hope it is okay to reask the question here.

$\endgroup$
  • $\begingroup$ why is the above C(p^2+q^2)/sqrt(npq)? $\endgroup$ – james black Apr 22 '18 at 6:51
6
+50
$\begingroup$

The paper "On The Bound of Proximity of the Binomial Distribution to the Normal One" - Nagaev, Chebotarev (2010) has improvements to $C$ specifically for the Binomial Distribution.

Theorem 2 on page 3 gathers together the results in the paper to show that $C$ can be taken to be .4215 in general. The paper notes that an Esseen (1956) paper demonstrates that the bound has to be at least $$\frac{\sqrt{10} + 3}{6\sqrt{2\pi}} \approx .409732$$ so they have found a relatively close bound.

However, there are some better bounds stated in the paper for specific situations. When $n \leq 2000$ and $.02 \leq p \leq .5$ the bound can be taken to be $.4096$. When $0 < p < .02$ it can be taken to be $.369$.

$\endgroup$

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.