2
$\begingroup$

Let $\ X_i \sim \mathrm{Pois}(10) $ be independent random variables $\ i = 1, 2,3,4,5 $ .

What is the probability of $\ X_1 = 9 $ given $\ \sum X_i = 45 $

now $\ \sum X_i \sim \mathrm{Pois}(50) $ and so I thought

$$\ P(X_1 = 9 \ | \ \sum_{i=1}^5 X_i = 45 ) = P(X_1 =9 \ | \ \sum_{i=2}^5X_i = 36) = \frac{P(X_1 =9 , \sum_{i=2}^5 X_i = 36)}{\sum_{i=1}^5 X_i = 45 } = \frac{e^{-5}\cdot \frac{5^9}{9!}\cdot e^{-40}\cdot\frac{40^{36}}{36!}}{e^{-50}\cdot\frac{50^{45}}{45!}} = \frac{0.03626 \cdot 0.053939}{0.0458} = 0.04269 $$

But apparently that is wrong and the solution is using binomial distribution with parameters $\ n = 45, p = 1/5 $ and final answer is $\ 0.14724 $

Is there anyway I can solve it with the Poisson distribution?

$\endgroup$
  • $\begingroup$ You can't change the condition to $\sum\limits_{i=2}^{5}X_i = 36$. However, it is true that $$P\left(X_1 = 9\, \Bigg{\lvert}\,\sum\limits_{i=1}^{5}X_i = 45\right) =\color{blue}{\frac{P\left(X_1 = 9, \sum\limits_{i=2}^{5}X_i = 36 \right)}{P\left(\sum\limits_{i=1}^{5}X_i = 45\right)}}$$ (this is different to $P\left(X_1 = 9\,\Bigg{\lvert} \,\sum\limits_{i=2}^{5}X_i = 36\right)$). What did you get when you tried to calculate that ratio of probabilities? $\endgroup$ – Minus One-Twelfth Mar 19 at 23:58
  • $\begingroup$ Thanks for the answer! I got $\ 0.0427 $ $\endgroup$ – bm1125 Mar 20 at 0:05
  • 1
    $\begingroup$ Maybe show us your working out that led to that answer? That way we can see if there was anything wrong with it. $\endgroup$ – Minus One-Twelfth Mar 20 at 0:06
  • 1
    $\begingroup$ Note that $5\times 5 =25$, not $50$. $\endgroup$ – Minus One-Twelfth Mar 20 at 0:14
  • 1
    $\begingroup$ OK, well, that changes your calculations, doesn't it? $\endgroup$ – Gerry Myerson Mar 20 at 0:16
1
$\begingroup$

It is because the sum of independent Poisson random variables is a Poisson random variables whose mean is the sum of their means, and all $X_i$ are independent Poisson random variables whose mean is $10$, that we have the following. $$X_1\sim\mathcal{Pois}(10)\\\sum_{i=2}^5X_i\sim\mathcal{Pois}(40)\\\sum_{i=1}^5X_i\sim\mathcal{Pois}(50)$$ (Emphasis because the fact of independence is crucial to this solution.)

Also, if we define $Y_\lambda\sim\mathcal {Pois}(\lambda)$ then we have $~\mathsf P(Y_\lambda{=}y)~=~\dfrac{\lambda^y\mathsf e^\lambda}{y!}\mathbf 1_{y\in\Bbb N}$, and we can use this to ensure we are substituting the correct values.

So, therefore, we get the following.

$$\begin{align}\mathsf P(X_1{=}9\mid \sum_{i=1}^5X_i{=}45)&=\dfrac{\mathsf P(X_1{=}9)\cdot\mathsf P(\sum_{i=2}^5 X_i{=}36)}{\mathsf P(\sum_{i=1}^5 X_i{=}45)}\\[1ex]&=\dfrac{\mathsf P(Y_{10}{=}9)\cdot\mathsf P(Y_{40}{=}36)}{\mathsf P(Y_{50}{=}45)}\\[1ex]&=\dfrac{\dfrac{10^9\mathsf e^{10}}{9!}\cdot\dfrac{40^{36}\mathsf e^{40}}{36!}}{\dfrac{50^{45}\mathsf e^{50}}{45!}}\\[1ex]&=\dbinom{45}9 \dfrac{10^9\cdot 40^{36}}{50^{45}}\end{align}$$

Which is where the binomial distribution comes from.

[The count of point events which occur in one among several independent Poisson processes has a conditionally binomial distribution when given the total count of events that occur among all of them, with success rate equal to the ratio of the sums of the Poisson rate parameters (favored versus total).]

$\endgroup$
1
$\begingroup$

$\newcommand{\bbx}[1]{\,\bbox[15px,border:1px groove navy]{\displaystyle{#1}}\,} \newcommand{\braces}[1]{\left\lbrace\,{#1}\,\right\rbrace} \newcommand{\bracks}[1]{\left\lbrack\,{#1}\,\right\rbrack} \newcommand{\dd}{\mathrm{d}} \newcommand{\ds}[1]{\displaystyle{#1}} \newcommand{\expo}[1]{\,\mathrm{e}^{#1}\,} \newcommand{\ic}{\mathrm{i}} \newcommand{\mc}[1]{\mathcal{#1}} \newcommand{\mrm}[1]{\mathrm{#1}} \newcommand{\pars}[1]{\left(\,{#1}\,\right)} \newcommand{\partiald}[3][]{\frac{\partial^{#1} #2}{\partial #3^{#1}}} \newcommand{\root}[2][]{\,\sqrt[#1]{\,{#2}\,}\,} \newcommand{\totald}[3][]{\frac{\mathrm{d}^{#1} #2}{\mathrm{d} #3^{#1}}} \newcommand{\verts}[1]{\left\vert\,{#1}\,\right\vert}$

Let $\ds{\mc{P}\pars{\lambda,n} = {\expo{-\lambda}\lambda^{n} \over n!}}$.

\begin{align} &\bbox[10px,#ffd]{\sum_{x_{1} = 0}^{\infty}\mc{P}\pars{\lambda,x_{1}} \!\!\sum_{x_{2} = 0}^{\infty}\mc{P}\pars{\lambda,x_{2}} \!\!\sum_{x_{3} = 0}^{\infty}\mc{P}\pars{\lambda,x_{3}} \!\!\sum_{x_{4} = 0}^{\infty}\mc{P}\pars{\lambda,x_{4}} \!\!\sum_{x_{5} = 0}^{\infty}\mc{P}\pars{\lambda,x_{5}}} \times \\[2mm] &\ \bracks{x_{1} = 9} \bracks{z^{45}}\! z^{\sum_{i = 1}^{5}x_{i}} \\[5mm] = &\ \mc{P}\pars{\lambda,9}\bracks{z^{45}}z^{9} \bracks{\sum_{x = 0}^{\infty}\mc{P}\pars{x}z^{x}}^{4} = \mc{P}\pars{\lambda,9}\bracks{z^{36}} \bracks{\sum_{x = 0}^{\infty}{\expo{-\lambda}\lambda^{x} \over x!}z^{x}}^{4} \\[5mm] = &\ \mc{P}\pars{\lambda,9}\expo{-4\lambda}\bracks{z^{36}} \bracks{\sum_{x = 0}^{\infty}{\pars{\lambda z}^{x} \over x!}}^{4} = \mc{P}\pars{\lambda,9}\expo{-4\lambda}\bracks{z^{36}} \expo{4\lambda z} \\[5mm] = &\ \mc{P}\pars{\lambda,9}\expo{-4\lambda}{\pars{4\lambda}^{36} \over 36!} = {\expo{-\lambda}\lambda^{9} \over 9!}\expo{-4\lambda}{4^{36}\lambda^{36} \over 36!} \\[5mm] = &\ \bbx{{4^{36} \over 9!\, 36!} \,\expo{-5\lambda}\lambda^{45}} \end{align}

$\ds{\lambda = 10 \implies \approx 6.7474 \times 10^{-3}}$.

$\endgroup$
  • 1
    $\begingroup$ Unfortunately, OP just edited the question, and wants Pois(10), not Pois(5). $\endgroup$ – Gerry Myerson Mar 20 at 0:17
  • 2
    $\begingroup$ $\Huge\left)\smile\right($ $\endgroup$ – Felix Marin Mar 20 at 0:20

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.