One could wish he had kept the focus on the qualitative question all the way through the passage you quote. "...then that is not consistent with things being only a little bit worse." But of course Dsquared has a highly individual rhetorical style, blunt and baroque at once, forthright and eccentric, and I wouldn't have it any other way.
Since it's already been linked at most of the liberal blogs today, I can only assume that this is part of your ongoing courtship of Mr. Davies.
2: But of course.
No, actually, I've been arguing about this at Obsidian Wings, and I thought Davies made a good point that pinpoint accuracy, and the breadth of the confidence interval, are really terribly irrelevant here.
LB, as I've pointed out more than once, Davies is not as fat and ugly as he makes it seem. His self-description made me think that he was a blimpish Den Beste of the left, sitting in his underwear in a room full of pizza boxes, but he's actually quite presentable.
For me, it's all about the vitriol.
At the end, he says:
But in the 2006 study, death certificates were checked and found in 92% of cases.
So how can the hospital & morgue counts be as far from the study estimate? Do the Iraqs only count one out of every ten bodies they issue certificates for? Or do they issue them the way my teachers used to issue study hall->library passes, by the whole pad?
Or am I missing something completely?
6 -- Comments to Kieran's CrookedTimber post advance some hypotheses about this.
Apparently the death certificate bureaucracy still functions, because people need death certificates for various very practical purposes, but the statistics-keeping division is trashed, because no one needs statistics. It was reasonably speculated that the death certificate people were taking bribes and thus had a motive for keeping things going.
Oops, I mean Daniel's CrookedTimber post. Though Kieran's is of course worth reading as well.
6: I'm going to dump this in here wholesale (it's from my blog) because it's fresh and it's relevant:
I get the sense that many people are accepting the new Lancet study as valid, or at least informative. But there’s still resistance, and notoriously that includes our own government. The resistors are happy to use inductive argument, and they seem to be settling on this:
Unlike the previous study, this new study cites death certificates as evidence. If the authorities in Iraq are able to issue death certificates, then the authorities in Iraq are also able to collate mortality data. Therefore, the mortality data issued by the Iraqi government is preferable to the figures from the new study.
You can object to this argument on the grounds that the issuing of death certificates in itself does not show that the Iraqi government is able to collate data. Death certificates are issued locally; data is collected centrally; communications in Iraq may not be good.
I think there are two additional possible responses.
One is to point out that the Lancet study bypasses any mechanism of statistical collection by the Iraqi government. The field researchers asked families for death certificates; in over 90% of cases, death certificates were shown. If the statistics produced by collation by the Iraqi authorities disagree with the Lancet study, and you want to challenge the study, scrutiny passes first to the mechanism of collation because you’re using that to build your case.
The other is to point out that the study does not depend on death certificates. It’s not enough to say that the death certificates were fakes; you also have to claim that the respondents were lying (and it’s certainly possible that a truthful respondent can be the holder of a fake certificate). Someone who pursues this line of argument will inevitably move very close to the position of claiming that Iraqi survey respondents lie by default.
Right. I have to admit that I'm not clear on the sourcing for what Davies says there, but the explanation that death certificates are issued by a whole bunch of decentralized sources (any doctor, etc.) and are then not centrally recorded in any effective way sounds reasonable.
10: Okay, I'm going with the notion that the Iraqi central (such as they are) authorities aren't working very hard at collecting the DCs and are quite happy to keep not working at it.
(And I'll agree, the precise number isn't very important. However, even when it comes to morality, orders of magnitude are.)
Right. That explanation requires a poorly working central statistical bureacracy, or fraud in the central bureacracy. The other explanation requires hundreds of independent frauds (for which it's not clear that there is any motive. The fact that a death certificate is useful when someone has died doesn't strongly suggest to me that it's useful when noone has.). The first seems much more likely.
By the way, everyone's talking as if the numbers coming out of the central statistical authority are much lower. Does anyone have a sense of what they are?
Jonathan Schwarz has a good post about orders of magnitude and morality.
Another way to look like a complete tool: when dealing with some statistical claim, imagine a possible world in which the data set looks just the same but the conclusion is far from true. Then appeal to this possible world as a refutation of the statistical claim.
The order of magnatude is the point. I mean, how many excess deaths would make the project worthwhile? I was also interested in the CT link about the comment re: Darfur having the same # of excess deaths. I would be in favor of an intervention in Darfur, but it is logistically not feasible. I think that 3/4 of what we decide to do is because we can, not whether it's right.
16: But the mortality in Darfur is happening through a lack of intervention. In Iraq, the intervention itself has produced the increase in mortality.
No argument with that.
I would be in favor of an intervention in Darfur, but it is logistically not feasible
Not trolling you, TLL, but why not? I would think that, worldwide, we can find enough people with helmets and guns and bags of wheat to deploy an overwhelming peacekeeping force. My sense is that we're simply not really trying very hard.
15 doesn't make any sense to me. Why does that make one a complete tool? If you read "statistical model" for "conclusion" (or "possible world"), then that's a pretty good description of a certain philosophy of statistical hypothesis checking -- Fisher-style testing with a null hypothesis, or something roughly along those lines. The words "data set look just the same" seems like a subtle appeal to some kind of inference based on likelihood.
Again, what gives? I'm not a Lancet Denier, but I just think 15 is kinda strange.
19: Yglesias and dsquared have discussed this issue a bit. Basically, the situation is a lot more complicated than is often assumed, and it's not clear where we (whoever that includes) are going to get all that stuff and whether it will help.
I didn't get that one either -- figured FL was sniping at some more specific behavior which he didn't adequately convey.
I should really be reading more Yglesias and dsquared. Got a specific link I can bum (or buy), teo?
Stanley- the reason an intervention is Darfur is not feasible is because we have no base to operate from and stage the intervention. My first assumption is that the government of Sudan would not welocme the interference. Darfur is some 900 miles from the Red Sea, over hostile territory then. That is a long airbridge, even if you decide to put a carrier offshore. The French still have some influence in Chad, so that is a possibility. But it's not like we have 150,,000 troops next door.
This Yglesias post has some links that are probably useful.
Thanks, teo and TLL. I, like many, have really had my head in the sand on the Darfur crisis.
a subtle appeal to some kind of inference based on likelihood.
It's just bayesing at the moon.
But the mortality in Darfur is happening through a lack of intervention.
Now that's just silly.
27: Omg, kiss my posterior. The expected value of that joke is l4me.
20: 'imagine a possible world' is philosophy-code for 'come up with any logically possible, but not necessarily plausible scenario, preferably involving cleverly painted mules and fake barns and brains-in-vats, and use that as a defeater against someone's argument. Then go drink.'
FL, I think, is referring to the sort of argument that would go, 'It's broadly logically possible that the Iraqis are lying to the reporters in an attempt to hasten the withdrawal of the United States, and that there's a huge secret shadow regime that instructs everyone to answer 'yes' when asked if they know personally someone who is killed, on the grounds that they all have TV and are awaiting the coming of the hidden imam. Also, we know the good news isn't reported, so they must have conducted their study only in the safe zones, which as we know always have more violence. Therefore, the study is discredited, because 12,000 is a small number.'
Brain in a Vat at the wheel of a runaway trolley.
The story should be updated to include uncultured brains in vats who smoke on trolleybusses.
Isn't being grown in a vat the very definition of a cultured brain?
If a cultured brain were smoking on a trolleybus that would be paradoxical indeed.
What particpant of high culture would deign to ride a trollybus?
An enthusiastic one! Not smoking!
13: Ironically, the Iraq Ministry of Health doesn't seem to provide figures based on the collation of regional mortality data. They do, however, seem to be providing the UN with a count of hospital deaths in Baghdad. Here's what the most recent Brookings Report has to say:
Information for May 2003-December 2005 is based upon data from Iraq Body Count. Our estimates since January 2006 are based upon the numbers published in the UN Assistance Mission for Iraq, “Human Rights Report: 1 May-30 June, 2006” and subsequent reports. This data combines the Iraq Ministry of Health’s tally of deaths counted at hospitals with the Baghdad Medico-Legal Institute’s tally of deaths counted at morgues. (The assistant director of the Baghdad morgue estimated that 90 percent of bodies at the morgue died from violence ...) ...
Combining IBC and two official sources of limited scope leads the Brookings Index to an estimate of 62,000 civilian deaths from violence since the invasion. This, presumably, is the 'order of magnitude' figure that the Blair team were talking about two days ago.
Emerson is using a lot of exclamation points this morning. Perhaps an intervention is in order.
Loafers lack enthusiasm. They do not live well.
But we wear our tuxedos on the beach. That's like totally Kul!