0
$\begingroup$

Consider the parameters $v_{a}$, $v_{b}$ be such that $0<v_{a}\leq v_{b}$ and $c>0$. I have an equation involving the Binomial distribution that I need to solve with respect to $p(n)$:

$\sum_{k=1}^{n}\left(\begin{array}{c} n-1\\ k-1 \end{array}\right)p^{k-1}\left(1-p\right)^{n-k}\cdot\left\{ v_{a}\left(\frac{1}{k}-\frac{k-1}{c\,k^{2}}\right)\right\} =\sum_{k=1}^{n}\left(\begin{array}{c} n-1\\ k-1 \end{array}\right)\left(1-p\right)^{k-1}p^{n-k}\cdot\left\{ v_{b}\left(\frac{1}{k}-\frac{k-1}{c\,k^{2}}\right)\right\} $

I can show that this solution has a unique solution $p\left(n\right)\in\left(0,1\right)$. (Assume that $c$ is such that each terms in brackets are positive for all $k$.) I wish to understand the solution to this equation, $p\left(n\right)$, in the limit $n\rightarrow\infty$.

I conjecture that $\lim_{n\rightarrow\infty}p\left(n\right)=\frac{1}{2}$. Any way to show this? Hint: The terms in brackets strictly decrease in $k$ and as $n$ grows the Binomial coefficients shift to the ``right'', so each of these sums should go to zero. Is this enough? I am concerned with the rate they go to zero... Thank you in advance for any inputs.

$\endgroup$
0
$\begingroup$

The answer to the question is that $\lim_{n\rightarrow\infty}p\left(n\right)=\frac{v}{v+1}$, where $v=\frac{v_{a}}{v_{b}}$. That is, my conjecture is correct, if and only if, $v_{a}=v_{b}$. (Existence and uniqueness) $p\left(\cdot\right)$ is bounded and strictly increasing. Hence, it converges to its supremum, which is unique. To explicitly find the limit $\lim_{n\rightarrow\infty}p\left(n\right)$, one has to use the fact that the LHS is the expectation of indenpendent non-IID $\text{Bin}\left(n,p\left(n\right)\right)$ and check that a suitable version of the CLT applies.

$\endgroup$

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.