Sunday, May 22, 2005

Lancet's Bogus Numbers.

Recently I had someone drop the bogus Lancet number of 100,000 civilian deaths in Iraq attributable to the invasion. I have never been convinced the numbers were true but really could not say why.

Today I spent some time looking at the critiques of the paper published by the Lancet. They strongest criticism center on the study itself. The researchers used faulty models to extrapolate their empirical findings. The criticisms also say self-reporting is not going to work here either. I would not be surprised if there is a fair amount of double counting of deaths in the study. Ali Abdulla knows two people killed in by an airstrike and so too does his brother.

Anyway the number on criticism I see is they surveyed 33 provinces (or localities) and throughout Iraq most are peaceful, but they oversampled the unpeaceful areas and then extrapolated that information to the peaceful areas. The report makes no distinction between directed lethality vs. random lethality. That is Fallujah is going to have a high number of deaths due to the war just because that is where the war was at, while Basra will have few fatalities. The study treats Basra and Fallujah alike. So if Fallujah found 10 ward dead per 1000 then the study assumes this to be true in Basra as well.

Also, there was pressure from the study's author to print his study prior to the election. Clearly this is politicization of science, and there is reason to doubt the study. Other studies do not have numbers as high, but they are not insignificant either.

See: The Command Post and The Chicago Boyz for more on this (these are the sites I spent most of the time reading in preparing to write this).