Also remember that the Lancet article was nearly six months ago. So with the 100,000 deaths figure covering the reported period calculated at 180 deaths per day, we should actually be talking about 130,000 deaths by now.

See, this is exactly the sort of fallacy I'm talking about.

"180 deaths per day" is a meaningful statement about a given period of time that has already been measured. It is not something that can be extrapolated, as it does not measure any inherent quality of the days concerned. It is a measurement of fact, not a rate of occurrence. Treating it as a rate is the sort of reasoning that leads people to claim that green cars are more dangerous than blue ones.

One of the big problems statisticians run into with the public is that people constantly misuse averages. If you take an average of the 100000 figure, you get 180 deaths per day, but, although this figure is useful for trying to envisage the scale of what's being described, it is not useful statistically, because the peaks and troughs in the curve are not random. The number of deaths each day is largely dependent on the tactics and rate of success of the armies and the terrorists, all of which change regularly. If you want to extrapolate the figure, you could cross-reference your findings with a timetable of tactical changes provided by the army, which would enable you to make reasonable estimates about the resultant death-rate of different tactics, and you might be able to make a start towards accuracy that way.

Here's a simple illustration of the problem. Take the total number of British deaths in WW2. Divide by the number of days in WW2. Multiply the result by the number of days from 1/6/40 to 1/7/40. Will that give you an accurate figure? No, not even close, because those sums fail to take account of the different levels of fighting during different periods of the war.

In case you're wondering, I'm not having this argument because I support the war. Like I said, I see no reason to disagree with the Lancet report's findings; I just disagree with the way that statisticians tend to interpret that type of finding, and, even more so, I get annoyed by people who misuse statistics (that's not "misuse" as in "use for a purpose I disagree with" but as in "manipulate without understanding the mathematics they're trying to use"). I don't believe that a war ceases to be right and magically becomes wrong when the number of casualties crosses a certain number, so the argument about the Lancet report is entirely academic to me.

Thanks Squander Two; it's nice to see something like this that isn't banging on for either 'side'.

As a scientist myself, if I'd come up with a 95% CI that ranged from 8000 to 198000, well the first think I would do is chuck the data out and start again. If I absolutely had to pull some kind of conclusion from it, I would probably say that the number of deaths is very likely to be at least 8,000, and leave it at that.

Hi, Joe.

Yup, that's what I'd do, too. That's why I agree with the Lancet report's findings: as far as I'm concerned, they found that there had been at least 8000 civilian deaths, which is plausible and unsurprising.