The number of accidents follows a Poisson distribution with mean $12$. Each accident generates $1,2,$ or $3$ claimants with probabilities $\frac{1}{2},\frac{1}{3},\frac{1}{6}$ respectively. Calculate the variance in the total number of claimants.

This is a SOA exam question so I have been provided the answer and they solve it like so:

Treat as three independent Poisson variables, corresponding to $1,2,$ or $3$ claimants

rate$_1 = 6$, rate$_2 = 4$, rate$_3 = 2$

Var$_1 = 6$, Var$_2 = 16$, Var$_3 = 18$

Total Var $= 6+16+18 = 40$

I understand where the numbers are coming from. The rates are the probabilities of the number of claimants times the mean of $12$ and the variances are the rates times the number of claimants squared.

However, what I don't understand is why we using those numbers? Why is variance calculated that way?

Also, could we possibly do this using Wald's Theorem instead and how would one do that?

  • $\begingroup$ Let $X$ be the number of one claimant accidents, and define $Y$ and $Z$ similarly. Then $X$, $Y$, and $Z$ are indeed Poisson with the rates given. The total number of claimants is $X+2Y+3Z$. If we assume that the random variables $X,Y,Z$ are independent, the variance of $X+2Y+3Z$ is as given. But I do not see that $X$, $Y$, and $Z$ are independent. $\endgroup$ – André Nicolas Feb 4 '16 at 3:33

Let $$N \sim \operatorname{Poisson}(\lambda = 12)$$ be the frequency distribution of the number of accidents. Let $X_i$ be the discrete probability distribution of the number of claimants in accident number $i$, with mass function $$\Pr[X_i = 1] = \frac{1}{2}, \quad \Pr[X_i = 2] = \frac{1}{3}, \quad \Pr[X_i = 3] = \frac{1}{6}.$$ Let $$S = X_1 + X_2 + \cdots + X_N$$ be the random total number of claimants. The question implies that $X_i$ and $X_j$ are independent whenever $i \ne j$. (If not, then the phrase "each accident generates 1, 2, or 3 claimants with probabilities $\frac{1}{2}$, $\frac{1}{3}$, and $\frac{1}{6}$, respectively" would not be sufficient to specify the probability distribution of $X_i$.) Then we have, by the law of total variance, $$\begin{align*} \operatorname{Var}[S] &= \operatorname{E}[\operatorname{Var}[S \mid N]] + \operatorname{Var}[\operatorname{E}[S \mid N]] \\ &= \operatorname{E}[N \operatorname{Var}[X_i]] + \operatorname{Var}[N \operatorname{E}[X_i]] \\ &= \operatorname{Var}[X_i] \operatorname{E}[N] + \operatorname{E}[X_i]^2 \operatorname{Var}[N]. \end{align*}$$ Since $$\operatorname{E}[N] = \operatorname{Var}[N] = 12,$$ and $$\operatorname{E}[X_i] = \frac{5}{3}, \quad \operatorname{Var}[X_i] = \frac{5}{9},$$ we have $$\operatorname{Var}[S] = 40.$$


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.