Expanding on this question: Conditional Probability Given N=n

Suppose that $N$ is a Poisson random variable with parameter $\mu$. Given $N=n$, random variables $X_1, ... X_n$ are independent with uniform (0,1) distribution. So there are a random number of $X$'s.

c) Let $S_n=X_1+...+X_N$ denote the sum of the random number of $X$'s (If $N=0$ then $S_N=0$). Find $P(S_N=0$).

I'm thinking: $$P(S_N=0)= P(S_N=0|N=0)\cdot P(N=0)$$ By condition $(N=0) \to (S_N=0):$ $P(S_N=0|N=0)=1$ so $P(S_N=0)$ is exponential distribution for $0$.

How can I carry this part forward to solve d) Find $E(S_N)$ $$E[E(S_N|N=n)]=E(pN)=pE(N)=p\mu$$


You have the law of total expectation: $$ \mathbb{E}[S_N]=\mathbb{E}[\mathbb{E}[S_N|N]] $$ Now, $$ \mathbb{E}[S_N|N] = \mathbb{E}[\sum_{n=1}^N X_n|N] = \sum_{n=1}^N \mathbb{E}[X_n] = \sum_{n=1}^N \frac{1}{2} = \frac{N}{2} $$ so $$ \mathbb{E}[S_N]=\mathbb{E}\left[\frac{N}{2}\right]=\frac{\mu}{2}\,. $$

Note: this is a special case of compound Poisson distribution.

  • $\begingroup$ We can simply use Wald's identity to compute $$ \mathbb E[S_N] = \mathbb E[N]\mathbb E[X_1] = \frac\mu2. $$ $\endgroup$ – Math1000 Nov 24 '19 at 0:21
  • 1
    $\begingroup$ Yes. But I didn't want to assume knowledge of Wald's identity here. It's useful to derive the result at least once by oneself... $\endgroup$ – Clement C. Nov 24 '19 at 0:22
  • $\begingroup$ I suppose the law of total expectation is more elementary. $\endgroup$ – Math1000 Nov 24 '19 at 0:23
  • 1
    $\begingroup$ Well, I'd say so. Without it, conditional expectations become quite unmanageable... $\endgroup$ – Clement C. Nov 24 '19 at 0:24
  • 1
    $\begingroup$ @ElliottdeLaunay $X_n$ is just a r.v. uniform on $(0,1)$. Its expectation is $1/2$. $\endgroup$ – Clement C. Nov 24 '19 at 1:21

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.