0
$\begingroup$

On page 4 of this article, the authors wants to find the normalizing constant $c$ but it is very hard to compute so I used the same formula to bound $c$.

Take the non-negative integers $~0 \leq x_1,x_2,...,x_r \leq X$ where $\sum_{i=1}^{r}x_i=X$. Using the the Multinomial distribution to find the probability of $x_1,x_2,...,x_r $ and assuming that $ p(x_i)=p=\frac{1}{r}$; where $r$ is a constant, then:

$$P(x_1,x_2,...,x_r;\frac{1}{r})= \frac{1}{c}\frac{X!}{x_1!x_2!...x_r!}\prod_{i=1}^{r} p^{x_i}$$. $$= \frac{1}{c}\frac{X!}{x_1!x_2!...x_r!}p^{\sum_{i=1}^{r}x_i}$$

To prove hat $c\leq1$, I did:

$$ c= \sum_{x_1+x_2+...x_r=X}\frac{X!}{x_1!x_2!...x_r!}p^{X}$$ $$c \leq \sum_{x_1+x_2+...x_r=X}\frac{1}{r^{X}}$$. let $y=\frac{1}{r}$, $$c \leq \sum_{0}^{X}{y^{X}}$$ $$c \leq\frac{1-\frac{1}{r^{X+1}}}{1-\frac{1}{r}}$$ $$c \leq \frac{r(r^{X+1}-1)}{r(r^{X+1}-r^{X})}$$ $$c \leq \frac{r^{X+1}-1}{r^{X+1}-r^{X}}$$

Is this correct or not? and if not, which part I missed?

P.S: Please write a comment if there is something unclear in the question so that I can clarify it or edit the question if needed.

Thanks for help.

$\endgroup$

1 Answer 1

2
$\begingroup$

The distribution $$ \mathbb{P}_{\mathbf{p}}(X_1=x_1,X_2=x_2,...,X_r=x_r)= \frac{1}{c}\frac{X!}{x_1!x_2!...x_r!}\prod_{i=1}^{r} p_i^{x_i} $$ is a standard Multinomial distribution. The normalising constant is thus $c=1$. Since $$ \sum_{\substack{x_1,\ldots,x_r\\\sum\limits_{i=1}^r x_i=X}} \frac{X!}{x_1!x_2!...x_r!}\prod_{i=1}^{r} p_i^{x_i} = \left\{\displaystyle{\sum_{i=1}^r} p_i\right\}^{X}=1\,. $$

$\endgroup$

Your Answer

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

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