If you don’t accept that the 100,000 number from the Lancet study on Iraq war causalties represents a probable lower bound (given its exclusion of Falluja, where we appear to have killed everyone) on the number of Iraqis who died in the 18 months following the war and otherwise wouldn’t have died in the 18 months following the war, and you do not have a PhD in a statistical discipline, then you are an ignorant bigot
Given this, the number of ignorant bigots out there is scary enough to make one lose the will to live. Until one remembers that (per Dan), probably over 100,000 Iraqis didn’t even get that option.
I’ve also lost the will for Christopher Hitchens to live, although that started to happen some time ago.
[*] See here for why. Respond with criticisms of Tim Lambert’s explanations only if you either have a PhD in a statistical discipline, or wish to provide my readers with cheap laughs at your ignorant bigotry.
"Respond with criticisms of Tim Lambert’s explanations only if you either have a PhD in a statistical discipline, or wish to provide my readers with cheap laughs at your ignorant bigotry."
I resent this sort of argument-from-authority. All I ask is that the people still arguing against that study please for the love of god just make the tiniest little effort to acquire some kind of common sense / intellectual rigour / grasp of arithmetic / other (delete as applicable) before doing so, or — vastly preferable — just shut the fuck up.
I’m sorry to say that I haven’t anything like Tim or Daniel’s patience to keep arguing against these fuckers when they carry on coming back with the same idiotic arguments again and again and again. The whole thing is pathetic, and the most tragic part of it — well, leaving aside the fact that, as you mention, we’ve killed tens of thousands of people in Fallujah and probably more than a hundred thousand more in the rest of Iraq for no reason — is that by repeating their shitty "arguments" these people have managed to turn the conclusions of a pretty good and very important study into "disputed facts".
It’s quite hard to find anything optimistic to say about this and I don’t think I have it in me to do so.
Are people without PhDs in statistics allowed to say that the study is correct?
Yes, as long as they can follow and cite the arguments used by the statistics experts who demolish the objections to the study.
(equally, should someone without extensive statistical knowledge know of someone with extensive statistical knowledge who has an unrefuted objection to the piece, they would do well to link it. I know of no person fitting the latter category, much as Tim Blair might fancy himself as an expert in anything other than Being A Cock).
I have some knowledge of epidemiological methods, and sampling, and have concerns with the study – especially with people who glibly use it as proof 100,000 people died. Does that make me a bigot?
I’d guess John will respond with "that depends on whether you say what the concerns are, rather than just hinting at your qualifications and saying that you ‘have concerns’". Certainly, that’s my reaction.
Anthony, I’ve asked you a couple of times to say what these concerns are and I think this is the nearest we’ve got. They seemed to be: 1) that I had said that inaccuracy of cluster sampling was an advantage 2) that epiemiological studies of hormone replacement therapy had given wrong results and 3) that some people who talked about the Lancet study sounded like Andrew Wakefield.
1) is actually not true, and 2) and 3) aren’t exactly knockdown criticisms. You don’t appear to have written anything about it on blacktriangle.org (or if you did, your post didn’t include the word "Lancet" which seems a bit perverse), and as far as I can tell, you’ve never commented on any of my, Chris’s or Tim Lambert’s Lancet stories. As I’ve mentioned in the past, I’d be really interested in what your concerns with the study are.
"that I had said that inaccuracy of cluster sampling was an advantage"
Cluster sampling is known to be less accurate method used in resource poor or difficult environments. Just because you state it is an advantage does not make it so. I do not understand why you have such a problem accepting that cluster sampling may not be an ideal sampling method, (there are better methods – which I accept may not have been possible to carry out at the time), while still accepting that study’s general points. Even good medical studies have accepted weaknesses.
It makes it look like you are defending the study as a truth, rather than as one study, which may, or may not, be an accurate portrayal of deaths in Iraq.
As regards not making a big issue of it on my site, there are a few reasons, one of which is I’ve not had time to cover it properly and would rather not do so glibly – as many who accept it, and defend it to the hilt, have done.
Note to Chris Lightfoot, the fact that the study was published in The Lancet (appeal to authority) does not make it a unassailable truth, as the Wakefield debacle shows, quite frankly the whole purpose of medical journals is to provide a forum to dispute "facts".
Personnally I put a plague on both your houses. Those who completely discount the study are as bad of those of you who say "Blair and Bush killed 100,000". Let’s not kid ourselves on this issue, for many it is just a stick to beat Blair with and a welcome distraction from some other uncomfortable facts.
I think Dan is claiming that he *didn’t* say the inaccuracy of cluster sampling was an advantage (and my response to your question is, indeed, roughly what Chris predicted.)
"Note to Chris Lightfoot, the fact that the study was published in The Lancet (appeal to authority) does not make it a unassailable truth…"
I couldn’t give a fuck where it was published. My opinion of it — which is not, as you imply, that it is an "unassailable truth" — is based on having read it.
Purely out of interest, why do you think it is useful to bring up this sort of straw-man nonsense here?
10,000 or 100,000 or 1,000,000 deaths. What difference does it really make? It was either a good decision or a bad decision to go to war. There isn’t some magic, tipping-point number of deaths that would turn it from being a good thing into being a bad thing. Even if there was such a number, no-one would agree on what it is. Everyone has already made up their mind as to whether it was a ‘good’ war or a ‘bad’ war already.
> Yes, as long as they can follow and cite the arguments used by the statistics experts who demolish the objections to the study.
Bollocks. You didn’t post an objection to people who trust scientific reports without understanding them; you didn’t post an objection to people who argue either for or against scientific reports without understanding them: no, the whole point of your post was to call anyone who disagrees with the report without understanding it an ignorant bigot. If you have just as strong an objection to those who blindly agree with the report, why not mention it?
I’m all for telling people to learn science or shut up about it, but it cuts both ways. For instance, I’m sick to the back teeth of people who ridicule the ignorance of Creationists without having the faintest clue about the theory of evolution other than "It must be true because scientists say so." Which is most people.
I had a huge comment to post here, but then I remembered that not clogging up other blogs’ comments was one of the reasons I started blogging.
The reason is that (as with evolution), the overwhelming weight of evidence is in the report’s favour, and I’m far more averse to people who are ignorant and wrong than people who are ignorant and right, annoying though the latter can be.
Here you go.
> the overwhelming weight of evidence is in the report’s favour
… which is immaterial to people who don’t understand that evidence.
Yes, but I prefer (taking a deliberately silly and extreme case) people who doesn’t axe-murder because God tells them that axe-murdering is wrong to people who do axe-murder, even though I believe that "because God tells you X" is a ridiculous reason to hold a belief.
Just because you state it is an advantage does not make it so
Just because you state that I state this (here and at Harry’s Place) doesn’t mean that I do. I’ve told you this twice now.
Anthony, if you don’t want to comment on the subject "glibly", then there is always the option of not commenting on it at all. To show up every now and then saying "I have all sorts of concerns with that study, but I’m not going to tell you any specific ones" looks something worse than glib. If all you’ve got an argument against is the claim "100,000 civilians dead" then say so. But you appear to be claiming, nonspecifically, that you think that there is reason to doubt the central result of a mortality rate rising by 50%. There’s a good old English, five words, blank blank or blank blank, that comes to mind here.
One of the problems I have is that pre-war regime-led violence was under-estimated by the study due to it’s sampling method and time-frame – unless one is willing to discount the deaths of hundreds of thousands of Saddam’s victims as a historical occurence with no relevance to the future conduct of the regime (or a successor regime involving his sons).
For example, one cannot merely say 100,000 deaths means this war was wrong, as some do, when things like this were occuring in Iraq before the invasion. Part of this, I admit is less a criticism of the study itself than uncomfortable feeling I have that some people are using this study to justify their opposition to the war (and the obvious point that this would have led to Saddam remaining in power). The study cannot give the full balance of benefits and harms of the war – which still remain to be seen. Still less the paper cannot address the benefits and harms of not having the war – which would have included continuing repression, torture and murder of the population.
Certainly a violent civil war and regime change in Iraq by the population without outside help would have been unlikely to have been less bloody than the Coalition invasion.
Excuse grocer’s apostrophe!
"One of the problems I have is that pre-war regime-led violence was under-estimated by the study due to [its] sampling method and time-frame…"
i.e., they measured the death rate in times [t1, t2] and [t3, t4] but you would — on grounds that you don’t specify but appear to have something to do with personal political views — much rather that they had instead measured the death rates in [t1′, t2′] and [t3, t4].
But that’s not an argument about the methodology of the study; it’s an argument about your political views. I assume when you said that you had "concerns with the study" you meant something more substantive and relevant?
Look, Anthony, if your position is that the deaths in 1989 and 1991 are the justification for a war in 2003 (I think that this is – part of – Norman Geras’ view), then that’s a defensible position. In my view, it’s wrong, but it’s defensible.
But it’s clearly not a methodological criticism of the study. Can we get this one cleared up please?
I’ve commented here. I think the Rogers figure is too high when compared to other conflicts, and I think many of the interviewees may have misrepresented or exaggerated their losses.
Perhaps, but I think you’re missing an explanation why in 81% of the cases where the researchers asked for death certificates, they were produced and in the remaining cases there were good reasons produced why there was no death certificate.
By the way, I don’t understand why you’re claiming that the death rate found by the Lancey study in Iraq is comparable to that in Darfur; it’s not. 100,000 deaths in eighteen months out of a population of c20m Iraqis is a much lower rate than ten thousand deaths a month out of c6m Darfurians
If we can’t even get an accurate estimate of the number of demonstrators at the Stop the War Coalition march yesterday in central London (was it 20,000 or 45,000 or 100,000 or 200,000) then how on earth do we expect to get accurate figures from Iraq on those who have died?
More seriously, the actual figure of Iraq dead is important since if the 100,000 or more figure is true, it would be a devastating blow to those who saw the invasion as the liberation of Iraq. My guess is that before invasion, the coalition estimated civilian casualties in Iraq would be at most a very few thousand. My feeling is that a pre-invasion estimated figure of 100,000 dead would have put the brakes on a coalition invasion.
I do have experience with statistics; I cannot see anything wrong with the study as such, but there is the possibility of a very large degree of error in such extrapolatory statistics. A major assumption is that those interviewed were telling the truth, as there was no confirmatory data as far as I can see. Also, the death rate would suggest also many more hundreds of thousands injured, and I am puzzled as to why figures of such large numbers of wounded were not also collected.
My gut feeling is that the estimate is far too high. However, as with all studies based on statistics, what is urgently needed are further studies on deaths in Iraq. Only then will a reliable death toll be possible. The figure of 100,000 is just not reliable enough at present to be given real weight. However, from the anti-war literature, it seems that this figure is accepted as true.
as there was no confirmatory data as far as I can see
Apart from the death certificates?
"If we can’t even get an accurate estimate of the number of demonstrators at the Stop the War Coalition march yesterday in central London…."
Actually, if we cared enough, it’d be pretty easy to do so. Counts based on sampling from aerial photographs ought to be pretty accurate. See, e.g., these photographs from an anti-Iraq-war rally in San Francisco in February 2003.
You’ve put your finger on it right there, Chris. The aerial photograph method is accurate because it involves actually counting the number of people. The figures given by police and protest organisers and news reporters are hopelessly inaccurate because they aren’t based on counting. The Lancet report was of the latter type.
Dsquared, on infant deaths, it said there usually weren’t death certificates for infant deaths – which does not mean, of course, that the claims were false, just that they could be. Also there was talk of reluctance to confirm non-infant deaths for fear of violence, so they settled on confirming a sample. I think an interviewer in a situation where violence is in the air might consciously or unconsciously seek confirmation from the more law-abiding looking interviewees.
I hope to return to the topic of figures for comparable conflicts, Darfur etc. later. Unfortunately not now, because I have to get to work.
S2: you’re exactly wrong on the protest-count issue. The coppers and protesters base their figures on counting, and they’re generally considered inaccurate; the aerial view method is based on sampling (like the Lancet study), and is generally considered accurate.
"The aerial photograph method is accurate because it involves actually counting the number of people." — no — sampling and extrapolating, just like the Lancet study, actually. (Two questions: (1) did you not read the page? (2) If not, did you really believe that somebody sat down with a bunch of photographs counting and ticking off 65,000 people?)
Sorry, I worded that appallingly, and deserved to be verbally slapped for it. Apologies. In my defense, I was in the middle of putting up shelves, which is distracting.
The aerial-photo method is based on counting the number of people on, as you say, a sample of the photos, and extrapolating the rest. However, that initial counting is still quite accurate because you can see each person on the photo. The police and organisers base their counts on estimates of how many people are probably in a given area based on estimates of crowd density based on looking at a load of moving people. Both methods involve extrapolation, but the former extrapolates from a count and the latter from a bunch of estimates.
As far as I can see, this means that I don’t actually have a point and should shut up.
I think the Lancets study is true crap really.
all self proclaimed statisticians sophisticated comments notwithstanding. sorry.
there is no indication from les roberts report that any effort was done in getting objective results.
only smooth talk, no actual filtering on pollers bias, answerers bias etc.
if pollers came at my door in 2004, living in Iraq,-> I would not open the door (the pollers in question had a 90% success rate in getting access: carpet sellers all over the world take note!)
if they had their foot in the door, I would give them the answer I thought they would like.
this is what the sample, and the report got : a couple of thousand of anxious people giving the answer they thought was the best to get the pollers away from their porche.
some kind of study…why not count deads in the graveyard or some other objective facts? -> because that would have been verifiable , accountable work.
not suitable for quick president campaigning.
this report belongs in the same intellectual category as deciding to shoot 4 poor geese out of the air.