Michelle Malkin has the text of a paper criticizing the 2004 Lancet study. In short, Lancet's statistical work was fudged (by omitting data collected in Fallujah) in order to present results which were statistically significant. Using the full data set, their analysis falls apart because of statistical noise. This is not an unusual problem for many types of social analysis. The statistical noise is high enough that you need a huge sample size to make valid decisions.
On a personal note, hypothesis testing is cool. I've had to do it a few times for work and it is amazingly powerful when you are trying to identify problems.
Subscribe to:
Post Comments (Atom)
No comments:
Post a Comment