Tuesday, November 02, 2004

The 100,000 lie

In Scoop, of course, the anti-war movement has predictably been quick to misrepresent and exploit the Lancet study on Iraqi civilian deaths -
Findings indicated 100,000 more Iraqis died than would have been expected to die,...
Even if one accepts the methodology of the study as being reliable, this is not what the study said at all. It estimates that somewhere between 8000 and 194000 deaths occurred. If the statistics for Falluja are included then there is the possibility that the civilian mortality rate was actually lower during and after the war as compared to before.

But we now have the anti-war movement claiming the 100,000 deaths to be a fact. It's called lying.


Blogger Greyshade said...

I'm fascinated as to how a sound statistical estimate with a high but reported uncertainty has progressed from a report to a myth to a lie. Excluding Fallujah data gives an underestimate of the true value (we don't know how bad an underestimate because we don't know how many similar hotspots there are). The most likely figure for this UNDERESTIMATE is 100,000 although a considerably lower (or higher) value is quite possible.

The estimate with Fallujah gives a higher but broader confidence interval which includes negative numbers at its extreme low end but only if we assume a gaussian distribution which the data clearly doesn't support. Specifically the gaussian CI assumes that there should be as many low outliers as high but to match Fallujah a low outlier would need a negative absolute death rate which is physically impossible. If (like me) you think the difference between 50,000, 100,000 or 200,000 matters you should advocate more and bigger studies like this to improve the precision.

2 November 2004 at 6:00 PM  

Post a Comment

<< Home