How Lancet did it this time, according to Strategy Page.

The article has rather missed the point. If you read the Lancet study the point is the 650,000 is total estimated deaths in Iraq from all causes including natural.

The key issue is that this is something like 5 times the population death rate prior to the invasion even with Saddam in charge.

The methodology is pretty widely used for this sort of thing and has pretty widely agreed statistical errors which the study outlines.

The article has rather missed the point. If you read the Lancet study the point is the 650,000 is total estimated deaths in Iraq from all causes including natural.

No. It's the estimate of excess deaths over the preinvasion rate for a four year period. 600,000 of those are estimated to be excess violent deaths.

The key issue is that this is something like 5 times the population death rate prior to the invasion even with Saddam in charge.

No. It would be a little more than double the rate assuming the preinvasion death rate was as low as claimed.

The methodology is pretty widely used for this sort of thing and has pretty widely agreed statistical errors which the study outlines.

Widely used or even best possible method doesn't mean it is accurate or even applied correctly. I notice that the Lancet paper ignore internal fraud and external deception. These are issues that matter when your study could be subverted by a powerful intelligence agency. Also, apparently, the study has an unusually small number of clusters compared to similar studies. And we still have the problem that this estimate disagrees substantially with studies that supposedly use a similar approach and had a much larger sample size.

Heck, even the guy running Iraq Body Count, who by no means could be said to want to cut the coalition any slack, disagrees with those numbers, which are an order of magnitude larger than the high end of his count.

Strange, I've read several analysis of this which don't match your numbers Karl.

In particular I've seen the numbers reviewed at around 15 per 1000 now, versus 3 per thousand before.

Likewise I've also seen statisticians saying that the survey size is perfectly adequate for this type of survey and the error margins were in line with standards and the sample size.

The Iraq body count uses a pretty crude metric which is also flawed to be low as was the UN study.

As I've said elsewhere here, I suspect the number is in the middle, which is the pre-invasion rates are true probably mean the death rate has more than doubled.