Skip to main content
added 105 characters in body
Source Link
Brendan McKay
  • 37.7k
  • 3
  • 80
  • 147

In my paper "On Littlewood's estimate for the binomial distribution", Adv. Appl. Prob., 21 (1989) 475-478, copy at http://cs.anu.edu.au/~bdm/papers/littlewood2.pdf , I find sharp exact bounds on this sum. Taking $p=1/2$ in Theorem 2 gives:

$$B(k; n,1/2) = \sigma \cdot b(k-1,n-1,{1/ 2})\cdot Y({k-n/2\over \sigma})\cdot \exp({E(k; n,1/2)\over \sigma})$$

where:

  • $b(k-1; n-1,1/2) = {1\over 2^{n-1}}{n-1\choose k-1}$
  • $B(k; n,1/2) = \sum_{j=k}^n b(j; n,1/2) = {1\over 2^n}\sum_{j=k}^n {n\choose j}$
  • $Y(x) = Q(x)/\phi(x)$, where:
    • $\phi(x) := {1\over \sqrt{2\pi}} e^{-x^2/2}$
    • $Q(x) := \int_{x}^\infty \phi(u) du $
  • $E(k; n,1/2)$ is the error term, which lies between 0 and $\min(\sqrt{\pi/8}, {\sigma/(k-n/2)})$.
  • $\sigma = \sqrt{n}/2$.

The relative error is at most $O(n^{-1/2})$ for all $k$, better if $k$ is not close to $n/2$.

The above requires $\frac n2\le k\le n$. For $0\le k\lt \frac n2$, use $B(k;n,p) = 1 − B(n-k+1; n, 1-p)$.

Somewhere on the arXiv there is a paper making numerical comparisons of many such approximations. I can't find it just now, maybe someone else can.

In my paper "On Littlewood's estimate for the binomial distribution", Adv. Appl. Prob., 21 (1989) 475-478, copy at http://cs.anu.edu.au/~bdm/papers/littlewood2.pdf , I find sharp exact bounds on this sum. Taking $p=1/2$ in Theorem 2 gives:

$$B(k; n,1/2) = \sigma \cdot b(k-1,n-1,{1/ 2})\cdot Y({k-n/2\over \sigma})\cdot \exp({E(k; n,1/2)\over \sigma})$$

where:

  • $b(k-1; n-1,1/2) = {1\over 2^{n-1}}{n-1\choose k-1}$
  • $B(k; n,1/2) = \sum_{j=k}^n b(j; n,1/2) = {1\over 2^n}\sum_{j=k}^n {n\choose j}$
  • $Y(x) = Q(x)/\phi(x)$, where:
    • $\phi(x) := {1\over \sqrt{2\pi}} e^{-x^2/2}$
    • $Q(x) := \int_{x}^\infty \phi(u) du $
  • $E(k; n,1/2)$ is the error term, which lies between 0 and $\min(\sqrt{\pi/8}, {\sigma/(k-n/2)})$.
  • $\sigma = \sqrt{n}/2$.

The relative error is at most $O(n^{-1/2})$ for all $k$, better if $k$ is not close to $n/2$.

Somewhere on the arXiv there is a paper making numerical comparisons of many such approximations. I can't find it just now, maybe someone else can.

In my paper "On Littlewood's estimate for the binomial distribution", Adv. Appl. Prob., 21 (1989) 475-478, copy at http://cs.anu.edu.au/~bdm/papers/littlewood2.pdf , I find sharp exact bounds on this sum. Taking $p=1/2$ in Theorem 2 gives:

$$B(k; n,1/2) = \sigma \cdot b(k-1,n-1,{1/ 2})\cdot Y({k-n/2\over \sigma})\cdot \exp({E(k; n,1/2)\over \sigma})$$

where:

  • $b(k-1; n-1,1/2) = {1\over 2^{n-1}}{n-1\choose k-1}$
  • $B(k; n,1/2) = \sum_{j=k}^n b(j; n,1/2) = {1\over 2^n}\sum_{j=k}^n {n\choose j}$
  • $Y(x) = Q(x)/\phi(x)$, where:
    • $\phi(x) := {1\over \sqrt{2\pi}} e^{-x^2/2}$
    • $Q(x) := \int_{x}^\infty \phi(u) du $
  • $E(k; n,1/2)$ is the error term, which lies between 0 and $\min(\sqrt{\pi/8}, {\sigma/(k-n/2)})$.
  • $\sigma = \sqrt{n}/2$.

The relative error is at most $O(n^{-1/2})$ for all $k$, better if $k$ is not close to $n/2$.

The above requires $\frac n2\le k\le n$. For $0\le k\lt \frac n2$, use $B(k;n,p) = 1 − B(n-k+1; n, 1-p)$.

Somewhere on the arXiv there is a paper making numerical comparisons of many such approximations. I can't find it just now, maybe someone else can.

Copy the relevant definitions from the linked paper, since in the paper they are sparsed over several pages.
Source Link

In my paper "On Littlewood's estimate for the binomial distribution", Adv. Appl. Prob., 21 (1989) 475-478, copy at http://cs.anu.edu.au/~bdm/papers/littlewood2.pdf , I find sharp exact bounds on this sum (take the case. Taking $p=1/2$ in ThmTheorem 2 for your problem).gives:

$$B(k; n,1/2) = \sigma \cdot b(k-1,n-1,{1/ 2})\cdot Y({k-n/2\over \sigma})\cdot \exp({E(k; n,1/2)\over \sigma})$$

where:

  • $b(k-1; n-1,1/2) = {1\over 2^{n-1}}{n-1\choose k-1}$
  • $B(k; n,1/2) = \sum_{j=k}^n b(j; n,1/2) = {1\over 2^n}\sum_{j=k}^n {n\choose j}$
  • $Y(x) = Q(x)/\phi(x)$, where:
    • $\phi(x) := {1\over \sqrt{2\pi}} e^{-x^2/2}$
    • $Q(x) := \int_{x}^\infty \phi(u) du $
  • $E(k; n,1/2)$ is the error term, which lies between 0 and $\min(\sqrt{\pi/8}, {\sigma/(k-n/2)})$.
  • $\sigma = \sqrt{n}/2$.

The relative error is at most $O(n^{-1/2})$ for all $k$, better if $k$ is not close to $n/2$.

Somewhere on the arXiv there is a paper making numerical comparisons of many such approximations. I can't find it just now, maybe someone else can.

In my paper "On Littlewood's estimate for the binomial distribution", Adv. Appl. Prob., 21 (1989) 475-478, copy at http://cs.anu.edu.au/~bdm/papers/littlewood2.pdf , I find sharp exact bounds on this sum (take the case $p=1/2$ in Thm 2 for your problem). The relative error is at most $O(n^{-1/2})$ for all $k$, better if $k$ is not close to $n/2$.

Somewhere on the arXiv there is a paper making numerical comparisons of many such approximations. I can't find it just now, maybe someone else can.

In my paper "On Littlewood's estimate for the binomial distribution", Adv. Appl. Prob., 21 (1989) 475-478, copy at http://cs.anu.edu.au/~bdm/papers/littlewood2.pdf , I find sharp exact bounds on this sum. Taking $p=1/2$ in Theorem 2 gives:

$$B(k; n,1/2) = \sigma \cdot b(k-1,n-1,{1/ 2})\cdot Y({k-n/2\over \sigma})\cdot \exp({E(k; n,1/2)\over \sigma})$$

where:

  • $b(k-1; n-1,1/2) = {1\over 2^{n-1}}{n-1\choose k-1}$
  • $B(k; n,1/2) = \sum_{j=k}^n b(j; n,1/2) = {1\over 2^n}\sum_{j=k}^n {n\choose j}$
  • $Y(x) = Q(x)/\phi(x)$, where:
    • $\phi(x) := {1\over \sqrt{2\pi}} e^{-x^2/2}$
    • $Q(x) := \int_{x}^\infty \phi(u) du $
  • $E(k; n,1/2)$ is the error term, which lies between 0 and $\min(\sqrt{\pi/8}, {\sigma/(k-n/2)})$.
  • $\sigma = \sqrt{n}/2$.

The relative error is at most $O(n^{-1/2})$ for all $k$, better if $k$ is not close to $n/2$.

Somewhere on the arXiv there is a paper making numerical comparisons of many such approximations. I can't find it just now, maybe someone else can.

Source Link
Brendan McKay
  • 37.7k
  • 3
  • 80
  • 147

In my paper "On Littlewood's estimate for the binomial distribution", Adv. Appl. Prob., 21 (1989) 475-478, copy at http://cs.anu.edu.au/~bdm/papers/littlewood2.pdf , I find sharp exact bounds on this sum (take the case $p=1/2$ in Thm 2 for your problem). The relative error is at most $O(n^{-1/2})$ for all $k$, better if $k$ is not close to $n/2$.

Somewhere on the arXiv there is a paper making numerical comparisons of many such approximations. I can't find it just now, maybe someone else can.