Can someone explain the mathematics of terrorism to me

[Wesley Clark]

http://www.physorg.com/news67524254.html

I don't get how an exponent of -2.5 really covers the level of terrorism as this implies a very low number of deadly attacks. Even a figure with one death per attack gives a number less than one. The PDF file talks about the formula p(x)=C*x^-a with a = 2.5. I know x is the value of the number of dead & injured in an insurgent attack, but don't know what C stands for.

And is this calculation per insurgent group or per insurgency?

I have no background in power laws, the only higher math I've taken is Calculus III.

Wikipedia implies

http://en.wikipedia.org/wiki/Power_law

That the C valus is the constant of proportionality, but what is that? Is that just a constant that is discovered by trial and error?

[ITR champion]

This article really looks fishy to me. It's saying that the number of attacks in any war can be determined by the power law relationship. If that were true, then any war would have to have one one-casualty attack per year. Doesn't seem right to me.

[Freddy the Pig]

Originally Posted by Wesley Clark

I don't get how an exponent of -2.5 really covers the level of terrorism as this implies a very low number of deadly attacks. Even a figure with one death per attack gives a number less than one. The PDF file talks about the formula p(x)=C*x^-a with a = 2.5. I know x is the value of the number of dead & injured in an insurgent attack, but don't know what C stands for.

C is the total number of "attacks" or "events" within a given war over a given period of time. The article doesn't say, but I presume an "event" is defined as an attack that has at least one fatality, because the formula doesn't appear to allow for events with zero fatalities.

All the formula does is tell you how the C events will be distributed in terms of number of fatalities. If there are 1,000 "events" in Iraq in the next year, we can expect that 1,000*2^-.25 = 177 will have two fatalities, 1,000*3^(-2.5) = 64 will have three fatalities, and so forth.

If you carry that out to infinity, you'll account for about 340 events, so I guess the other 660 must have one fatality. (If you apply the formula for x=1, you get p(x) = 1.000 which makes no sense.) The article doesn't really explain that point.

[DanBlather]

I'm too lazy to look at the article, but it sounds like a "long tail" distribution. Many different phenomena have the same sort of distribution: popularity of names, movie rentals, etc. The basic idea is that the most common event is very common, but that given enough events even uncommon ones will occur.