Saturday, March 19, 2005

Instapundit looks at the Lancet study claiming 100,000 deaths in Iraq, and I see someone has noted the authors made a basic mathematical mistake in claiming the data shows 98,000 is the most probable estimate.

That is a point I have made before, and I’m glad to see I’m not the only one who noticed: the distribution of the data inside the 95% confidence interval (8,000 to 194,000) is totally unknown; every number within it is equally probable. The equations simply don’t say anything about the interior of the confidence interval. It is a quanta, and nothing can be said about its structure because it has none as far as the confidence interval equation is concerned.

Grabbing that 98,000 number and saying it is more probable that any other number within the interval is simply wrong.

Only one statement can be made from this data (assuming you ignore/accept the methodology problems and author bias issues): there is a 95% likelihood the number killed was between 8,000 and 194,000. Any single number estimate made on the basis of this study is not an accurate representation of what the data actually says.


Post a Comment

<< Home