Suppose $\lambda > 0$ and $p \in (0, 1)$. The number of fish caught in one day follows a $\operatorname{Poi}(\lambda)$ distribution. At the end of the day, each fish is examined independently. With probability $p$, each fish passes inspection. Let $X$ be the number of fish that pass inspection. What is the moment-generating function of $X$?

The moment-generating function is given by $\operatorname E(e^{tX})$. I tried applying the law of total expectation knowing that the number of fish gives a partition of the space. This gives $$ \operatorname E (e^{tX}) = \sum_{i=0}^\infty \operatorname E (e^{tX} \mathbin{|} N = i) \operatorname P ( N = i ) $$ where $N \sim \operatorname{Poi} (\lambda) $ is the number of fish caught in the day. The expectation under the sum is then exactly the moment-generating function of a binomially distribution variable with number of trials $i$. So I get, $$ \sum_{i=0}^\infty \operatorname E (e^{tX} \mathbin{|} N = i) \operatorname P ( N = i ) = \sum_{i=0}^\infty (1 - p + pe^t)^i e^{-\lambda} \frac{\lambda^i}{i!} $$ but I'm not sure how to proceed in the evaluation of this series.

One idea I have is to combine the factors to the power $i$ under the sum to try turning the sum into a sum of the probability mass function of $\operatorname{Poi}(\lambda(1 - p + pe^t))$, which will go to one. \begin{align*} \sum_{i=0}^\infty (1 - p + pe^t)^i e^{-\lambda} \frac{\lambda^i}{i!} &= e^{-\lambda} \sum_{i=0}^\infty \frac{(\lambda(1-p+pe^t))^i}{i!} \cdot \frac{e^{-\lambda(1-p+pe^t)}}{e^{-\lambda(1-p+pe^t)}} \\ &= e^{-(1-p+pe^t)} \sum_{i=0}^\infty e^{-\lambda(1-p+pe^t)} \frac{(\lambda(1-p+pe^t))^i}{i!} \\ &= e^{-(1-p+pe^t)} \end{align*}

Is this procedure legitimate? It seems wrong to me as taking the first derivative of the moment-generating function and evaluating at $t=0$ gives $-p$, meaning that the expectation of $X$ would be negative!


I think I see some mistake here, you have $$e^{-\lambda}\cdot e^{\lambda(1-p+pe^t)}=e^{-(1-p+pe^t)},$$ while it should be $$e^{-\lambda(p-pe^t)}.$$

Then the first moment will be $\lambda p.$

  • $\begingroup$ You are absolutely right. It was merely a calculation error on my part. $\endgroup$ – Jacob Errington Apr 19 '17 at 21:08

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.