Calculating risk

#1
Hi

I'll start by saying that I'm a complete newbie and even though I've read many statistics books during the last 3 years, what I've learnt is very little.

My aim is to quantitatively measure risk, not the financial risk as it is usually done, but the fire risk (I work for an insurance company as a loss control engineer). Everybody uses heat maps that are completely useless.

Risk is defined by probability (some say is actually frequency) x severity and I'd like to develop a Monte Carlo simulation that can calculate that. In operational risk, banks have made many analysis and have come to the conclusion that severity obeys a lognormal distribution and probability (or frequency I'm not really sure) obeys a Poisson distribution. Do you think this is correct and could be applied to fire if I get to quantify probable losses?

What would be the parameters for the Poisson distribution then if the minimal value would be "0", the most common value would also be "0" (fires don't happen often fortunately) and the maximum value would be 3 or 4 for an extreme case of bad luck and controls.

So can anyone help me? Confirm or deny about the Poisson and Lognormal distributions and also how to write the formula in Excel?

Thanks a lot in advance.
 

hlsmith

Less is more. Stay pure. Stay poor.
#2
You need to make assumptions about the distributions of the two variables. I would look to the literature in your field for guidance. There is a distribution for "zero inflated Poisson". Yes, i would imagine this is achievable in Excel!