Let me reiterate - I couldn't care less if the risk of a tourist getting killed as a result of a terror attack is higher or lower in Egypt compared to the US or anywhere else.

However, if you're going to try and define that risk, either relative to other issues or to other locations, then use a meaningful model. To answer your question, I don't have the data to address the actual risk factor, and I have ZERO interest in searching it out on the web. What I can tell you is the model needs to be tailored to your risk assessment, and using Egypt as an example, you're better off separating Sinai from the rest of the country in terms of your variables (number of tourists, number of attacks, number of people killed, etc). Since you seem to have those numbers, run them and see what you get. If you insist on averaging, the anecdotal evidence seems to point to binomial distributions as opposed to normal ones.

OMNI had a thread this morning about crime statistics in NYC over the last 3 years (I think) with some interesting notes - quoting from memory roughly 50% of murder victims had a criminal record and a large % of murder victims were killed by family or friends (no correlation between the two, just large % in each case).

Regarding averages in general: GM in the early 90's was working with a small consulting firm on some market research issues and one of GM's marketing guys was going on about their average customer wanting this and saying that based on survey figures. The head of the consulting company, a woman, BTW, stands up and notes that the average GM customer has one tit and one ball, and how many customers like that have you seen lately?