See Tim Lambert's point here: the UN question refers to deaths from violence (33,000 of the total 100,000 Lancet toll), so the two studies match almost perfectly.

Bollocks they do John.

33,000 isn't in the confidence intervals of the new study and is a loooooong way from the 100,000 spouted as a fact routinely by some.

In fact, last time I worked it out 24,000 was 27% less than 33,000. The bigger the study, the more precise the point estimate and the narrower the confidence intervals.

Which leads me onto Squander point that:

One study says there were probably between 8,000 and 194,000 deaths; the other says there were probably between 18,000 and 29,000 deaths. There's obviously no contradiction there.

It isn't a contradiction. First you are not accounting for the fact the Lancet study included excess deaths from increases in disease, accidents and murders, so the actual number is lower.

Secondly, the fact that the two studies overlap at all means tyhat you cannot discount the fact that both studies could include the correct result. So look at this graphic. Compare trial 5 and trial 7. The correct number could be in either trial, but the bigger black square, representing the sample size, and the shorter confidence intervals, means you can be more confident that trial 7 is close to the truth than trial 5 (which has long confidence intervals and a smaller black box (indicating a smaller sample size).

Let's get this straight, I found the manner of the publication of The Lancet study highly suspect and Richard Horton's appearance on Newsnight made it clear that this was a highly political paper.

Meanwhile over on John's blog he's at the stage of calling people "statistical fuckwits who believe what they want to believe and lie to suit their own interests".

So sensible debate there then.

A 27% difference (although significant) is not relevant in the context we're talking about here, which is broadly "does the new study imply that the total death toll from the Iraq war is at the lowest end of the confidence intervals in the Lancet report?".
If the total Lancet figure is wrong by only 27% in either direction, the 100,000 figure is a reasonable rounding of whatever the correct number is.

A 400% difference, however, would have been highly relevant - and my invective was reserved for those who knowingly or should-have-been-knowing-ly twisted the figures to claim the latter. The writer of the Times piece is certainly one.

You remember Newman and Baddiel's sketches about Ray, the man afflicted with a sarcastic tone of voice? That's me, that is, only in text form. I actually had to deliberately develop a labouredly humourous vocal expression a few years ago because people kept believing me when I was talking total bollocks; now the pendulum seems to have swung in the other direction, especially in writing.

Anyway, Anthony, when I said "There's obviously no contradiction there" I meant that there was obviously no contradiction there. I agreed with every word you've typed before you typed them.


John,

> the UN question refers to deaths from violence (33,000 of the total 100,000 Lancet toll)

The Lancet did not give a figure of 100,000; it gave a figure of between 8,000 and 194,000. Now, I am well aware that there are all sorts of statistical reasons for choosing a probable figure of around 98,000 (although, as I said in my first post on this matter, I do not believe that statistical methods are much good at measuring this sort of thing). But that's not what you've done. You have asserted that 100,000 is a minimum, and you have repeatedly accused anyone who disagrees of being ignorant, stupid, biased, and bigotted. In that context, a 27% difference certainly is relevant.

Seeing as it's been quite a while since the Lancet article was published, and as the situation in Iraq has remained largely unchanged in the meantime, shouldn't those espousing the 100,000 figure be claiming more like 140,000 now ?

I seem to recall doing a brief calculation and coming up with 180 deaths/day, it must have been at least six months from the Lancet report by now.

> I seem to recall doing a brief calculation and coming up with 180 deaths/day

A brief and fundamentally flawed calculation, as I said at the time.

So the idea is that by quoting 100,000 deaths today, it is more and more likely to be an accurate estimate, whilst convieniently forgetting that the 100,000 deaths claimed some months back must have been false ?

I agree with what you are saying, but my argument is that surely the 100,000 deaths recorded by the Lancet article must be higher now if the situation that causes those deaths remains the same, I couldn't care less what the calculation is.

Argue as much as you like over 180/day, but you certainly can't support 0/day either, so why are they still claiming 100,000 deaths some six months later ?

Well, the situation hasn't remained the same, has it? The anti-warriors who keep quoting the 100,000 figure are complaining about deaths caused by Coalition forces, which is what happened during the main combat operations of the war. Most deaths now are being caused by the "insurgency", and the anti-warriors tend not to be interested in spreading anti-insurgency propaganda.


> the 100,000 deaths recorded by the Lancet article

No. The between 8,000 and 194,000 deaths recorded by the Lancet article.

Tsk. Here am I correcting incorrect statements, and I make one myself.

> The between 8,000 and 194,000 deaths recorded by the Lancet article.

No. The between 8,000 and 194,000 deaths estimated by the Lancet article.

The anti-warriors who keep quoting the 100,000 figure are complaining about deaths caused by Coalition forces ... Most deaths now are being caused by the "insurgency"

No, the real claim is more subtle than that. 100,000 people died in the time period covered by the survey that would likely not have died had Saddam been left in power. Even after accounting for Saddam's death squads.

Only some of those -24,000 or 33,000 - are due to direct action. Only some of those are due to insurgent violence. The rest are due to, well, anarchy and its effects on healthcare, clean water, law and order etc. The coalition isn't completely responsible for those, but bear at least part of the blame for failing to plan and manage the reconstruction effectively.

Yes, I'm aware of all that, but the people who repeatedly quote the 100,000 figure don't then follow it up with such qualifications and explanations. They're trying to give an impression that will encourage people to agree with them. If they talk about deaths caused during the war, it gives a clear impression that blame lies with Coalition forces. If they were to talk about deaths in, say, the last six months, it would not give that impression. That's all I was saying.


> failing to plan and manage the reconstruction effectively.

This gets repeated so much that people are beginning to think it's true. The reconstruction has been quicker and more efficient than anyone might reasonably have guessed. It's been far more impressive, despite having far greater obstacles in its way, than anything we've managed in, say, the UK or the USA. It wasn't so incredibly perfect that it had the entire country looking like France within a week, but I fail to see that that's a reasonable criticism.