March 17th, 2006

Про использование статистики - от любимого Малькольма Гладуэлла

This is a paradox that is not confined to dogs. For instance, for years people in the pharmaceutical business have been aware of the fact that a large number of reported adverse reactions to a particular drug can mean one of two things. The obvious meaning is that a drug is dangerous. The other meaning is that a drug is SO much better and safer and more effective than any other drug in its class that it tends to be given to the sickest and most troubled patients.

If, for example, a drug company company came up with the best anti-depressant in the world--something twice as good as Prozac--we would EXPECT that drug to be associated with, say, more reports of suicide ideation. Why? Because it would be prescribed overwhelmingly to the hardest cases, to the most depressed and suicide-prone sector of the psychiatric population.

The point is that we need to be very careful in the way we interpret statistics purporting to show that one kind of dog, or one kind of drug, or one kind of anything, is more dangerous than other things in its class.