3
$\begingroup$

Are there good lower/upper bounds for $ \sum\limits_{i = 0}^k {\left( \begin{array}{l} n \\ i \\ \end{array} \right)x^i } $ where $0<x<1$, $k \ll n$?

$\endgroup$
5
  • 1
    $\begingroup$ The ratio of two consequtive summands is $\frac{a_i}{a_{i-1}}=\frac{n-i+1}{i}x$ is unimodular. So you'll easily find the largest summand. And it is an upper bound for the whole sum (up to some constant), because another terms decrease not slower than geometric progression. $\endgroup$ Jan 6, 2018 at 10:01
  • $\begingroup$ BTW you can typeset binomial coefficient as \binom ni $\binom ni$. If you want bigger size, you can use \dbinom ni $\dbinom ni`$. (But I would not recommend the latter in the title. $\endgroup$ Jan 6, 2018 at 10:06
  • $\begingroup$ @AlexeyUstinov I doubt about geometric progression. Say, if $x=k/n$, the sum is approximately 1/2, but each summand is much less. $\endgroup$ Jan 6, 2018 at 10:19
  • $\begingroup$ (Sorry, I was thinking about a different sum $\sum \binom{n}i x^i (1-x)^{n-i}$, which is reduced to this sum and vice versa via introducying $x/(1-x)$ as a new variable). $\endgroup$ Jan 6, 2018 at 10:26
  • $\begingroup$ related to :mathoverflow.net/q/55585/51189 $\endgroup$ Jan 6, 2018 at 12:40

3 Answers 3

7
$\begingroup$

Write $x=p/(1-p)$ and then $$ \sum_{i=0}^k \binom ni x^i = (1-p)^{-n}\sum_{i=0}^k \binom ni p^k(1-p)^{n-k}.$$ The last sum is the cumulative binomial distribution, which has no exact formula (except as a special function) but a large literature on bounds. It is quite a common topic on Mathoverflow, see these for example: ref1 ref2 ref3 ref4 ref5

$\endgroup$
2
$\begingroup$

Let $p=\frac{x}{1+x}$ and $q=\frac{1}{1+x}$, and thus $$\sum_{i=0}^k \binom{n}{i} x^i=(1+x)^n\sum_{i=n-k}^n \binom{n}{i} p^{n-i} q^i.$$ Then for $k<np$ Chernoff bound gives $$\sum_{i=n-k}^n \binom{n}{i} p^{n-i} q^i \le \left( \frac{nq}{n-k}\right)^{n-k} e^{np-k}.$$ That is, $$\sum_{i=0}^k \binom{n}{i} x^i \le (1+x)^k \left( \frac{n}{n-k}\right)^{n-k} e^{\frac{(n-k)x-k}{1+x}}.$$

$\endgroup$
0
$\begingroup$

Hint :for lower Bound we have for $k> 1$:$ (1+\frac{1}{k})^k \leq (e^{1/k})^k =e ,0<x=1/k<1$ , and $1/k << k $

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

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