I think it is important to address at least one troublesome statistical bit about the catchy title. Obviously he put it up to get the attention American audience, but still some statistical problems get ignored in the whole article. I will tackle just one.
It is key to understand that small samples yield extreme results more often than large samples do which means the same as large samples are more precises than small samples.
A quick look at the last table, estimated fighters per million Muslims shows that the smallest countries tend indeed to give the most extreme results while the large countries tend to cluster around the center. Italy is at the first glance really the only outlier. (Interestingly the topic was and is pretty nonexistent in the Italian media)
Wikipedia gives a rough overview on muslim numbers per country:
Austria 475,000
Belgium 638,000
Bosnia 1,564,000
France 4,704,000
Denmark 226,000
Germany 4,119,000
Italy 1,583,000
Norway 144,000
Netherlands 914,000
Spain 1,021,000
Sweden 451,000
UK 2,869,000
The little I have read is that allmost all of the fighters stem from the male age group between 20-34, which usually makes up roughly 15% of the population. Even if we up this to 25% considering the immigration background to be on the safe side some numbers get very small indeed.
In Norways and Denmarks case we are around a pool of just 35000 and 55000 which make extreme outcomes of such rare events very likely. So no surprise that they are very far away from the mean.
This goes in doubly so for intelligence work, as they have to estimate their numbers for the whole country based on relative rare information about rare events.
This is a very tricky&hard thing to do which very likely will lead to considerable differences between the estimation process between countries even if would not take the chance factor in the relative rare informations about relative rare events into account. Even very smart guys in smart organizations will differ. Actually you can see that already to a good degree on their estimates. Some put an absolute number there, sometimes a round one, sometimes even an odd one, others have a bracket but the spread goes from roughly a 1/3 to factor 3!
To come back to the title it should be now obvious why it is a bad idea to take an extreme outcome from a small sample and to use it as base to get a 'perspective' for a nation the size of the USA.
The approach take in this short post is mostly based on Kahnemans work on 'small numbers'.
Reply With Quote
Bookmarks