Here is another question I have while practicing for the P/! exam:

An automobile policy owner has a 10% chance of having an accident in a $3$ year period. Given that an accident occurs, the payout for the accident follows the following distribution

$$f(x) = 0.5e^{-0.5x}$$

with $x$ representing thousands of dollars. If no accident takes place, the driver receives a rebate of $\$100$. What is the standard deviation of the payout?

Since $x$ appears to be following an exponential distribution, I thought that $\mu$ would be $2$ (or since $x$ is in thousands of dollars, the average payout to be $\$2000$ assuming there is an accident)

I know form the answer key that the answer is $\$851$, but I'm stuck as to why.

  • 1
    $\begingroup$ You have pretty much answered your own question: "the average payout [is] $2000 assuming there is an accident". But you also have to allow for the case in which there isn't an accident. $\endgroup$
    – David
    Jan 5 '18 at 0:38
  • $\begingroup$ David, my plan was to determine the variance then take the square root. Here's exactly how far I got: E(x) = .1($2000) + .9($100) = $290. However, when I go to calculate E(x^2), that's where I'm getting stuck. I thought to use the formula E(x^2)= 0.1(E((x_a)^2)) + 0.9(E((x_n)^2)) with x_a being the payout assuming there is an accident and x_n being the payout assuming there is no accident. $\endgroup$
    – John Capps
    Jan 5 '18 at 0:53

You have calculated the mean assuming there is an accident as $\$2000$. You could also calculate the standard deviation as $\$2000$ assuming there is an accident, making $E[X\mid A]=2000$ and $E[X^2\mid A]=2000^2+2000^2=8000000$

Similarly you could calculate $E[X\mid A^c]=100$ and $E[X^2\mid A^c]=0^2+100^2=10000$

That then makes

  • $E[X]=0.1\times 2000 + 0.9\times 100=\$290$
  • $E[X^2]=0.1\times 8000000 + 0.9\times 10000=809000$
  • $\text{Var}(X)=809000-290^2=724900$
  • the overall standard deviation $\sqrt{724900}\approx \$851$

In this case, if accident then the payout has mean $2000$ with variance $2000^2$. If no accident then the payout has mean $100$ with variance $0$. By the law of total variance, \begin{align*} \text{total var} &= \text{mean}(\text{var}) + \text{var}(\text{mean})\\ &= 0.1 \times 2000^2 + (0.9 \times 100^2 + 0.1 \times 2000^2 - 290^2)\\ &= 724900, \end{align*} and $\sqrt{724900} = 851.41$.

I just learned this in a probability class. I only vaguely understand it conceptually.

  • $\begingroup$ Thank you! I had thought to try to set it up using this formula, but by the time I thought of it, I was so frustrated with this problem and so tired that I couldn't figure out how to set it up. $\endgroup$
    – John Capps
    Jan 5 '18 at 12:42

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.