Given that the Lancet’s estimate is for the excess deaths to a time over a year ago, that the Iraqi infrastructure is still a mess, and that the civil war is ongoing, do you not think we may have reached the magic 100,000 already?

There’s nothing wrong with waiting for facts to come in before changing a position. Given how many facts we are already aware of, I can’t help wondering if I could come back to you in another 30 years for an answer, and hear “Not yet, not yet…”

…the Lancet paper does not say that the investigators wanted to exclude the most dangerous areas—-they said they wanted to reduce their total travel time, because it was dangerous to travel in Iraq.Okay. This does not diminish the argument that the choice of clusters that were grouped together was not random. The right way to reduce total miles traveled would have been to increase the size of the initial clusters, yielding fewer than 33 clusters. As it is, they paired several of the clusters based on a prior bias of estimated violence. Unfortunately, this reduces the veracity of the results.

If the Lancet investigators had done what you suggest, they would have biased their survey towards a lower death toll. They didn’t, although they did throw out their Fallujah data because the death toll was so high.

The authors wanted to exclude areas that they deemed too dangerous to travel to, but, of course, they paired thoses Governorates with places they could travel to, so, by definition, they were actually deemed less dangerous than the other half of the pair.

Any time a subjective judgement is put into the pairing of clusters, the random aspect is gone. Now, it could be that the authors were lucky in their picks and the pairing didn’t make much difference in the final outcome of the study.

In any case, the study appears to be quite weak and it’s not wise to mount policy decisions based on it.

It’s also my impression that there have been similar problems in other wars. If you wanted actual bodycounts of civilians killed by insurgents in the French/Algerian war, you can find fairly detailed records, because that’s the sort of thing the French government would keep track of and pass on to reporters. If you want bodycounts of Algerian civilians killed by French air raids and air strikes and infantry in remote villages, you can forget about exact bodycounts. You have rough estimates alone and if you insisted on confirmed numbers of dead actually counted by reporters or found in morgues or in French government reports, you would end up with a huge underestimate. If you read Alistair Horne’s “A Savage War of Peace”, his history of the Algerian conflict, you’d find detailed numbers given for those deaths that the French government would naturally keep accurate numbers for and would be willing to dispense—when he talks about the number of Algerians killed by the French it all goes hazy and he is reduced to citing various estimates made by different groups.

I would very much like to know the answer to that question.

I don’t, and don’t pretend to.

Or I could ask you if there’s any level of excess death that you think could make the invasion a bad idea?

You could, and the answer is obviously yes that there would be one that makes it unambigously a bad thing, rather than debatable as at present.

In fact, if the commonly quoted ‘100,000 casualties’ figure were shown to be factually true, then I certainly wouldn’t see much ambiguity about the issue.

If the civil war continues for another year, or another five years, will you keep putting on that happy face?

A year maybe, 5 years no.

Is the idea that anyone might let facts influence their judgement in any way completely outside of you experience?

soru

Given the particularities of this genre, it seems there should be, and I think there are, conventions that indicate how the Lancet article should be evaluated, and how the results of the evaluation should be presented. It seems odd that (as far as know) there have been no reviews that engage the article at its own level. Instead, the has been punditry and blogwriting, which have very different standards of argumentation, epistemology, and civility, certainly more permissive and less rigorous. (The merits of these standards in there own spheres is another subject, though I would not wish to see blogs and pundits adopting the style and substance of professional journals.) I’m not saying that the popular commentators have no business talking about scientific writing, that their criticism is inherently presumptuous, or that they should not loiter in respectable neighborhoods but keep to their own side of the tracks. Rather, popular essays are not the tools to use for the task of demolishing a scientific article. I hasten to say that there will be exceptions. That the argument about the Lancet study continues on essentially the same terms on which began a year ago suggests that no such exceptions have occurred in this debate, and none should be expected.

I may have a misguided view of science (never my strong suit). Yet I think that there are knowledgeable persons (or institutions) who could evaluate the Lancet study in light of applicable theories and accepted methods, present their assessment in a form suitable to the project, and that the effort would not be unusually difficult as compared to reviews of other scholarly articles. Such reviews, I am increasing convinced, occur on a regular basis, and in general contribute to the advancement of science and knowledge. Who will rise to this occasion?

Or I could just ask you how the mortality rate now compares to the mortality rate in 2000.

Or I could ask you if there’s any level of excess death that you think could make the invasion a bad idea? If the civil war continues for another year, or another five years, will you keep putting on that happy face?

Surely the situation in Iraq is best split up into 3 eras:

1. pre-war: saddam in charge of country, leaky sanctions, coalition planes occasionally bombing uniformed iraqi forces, ongoing state terror.

2. war: Saddam in charge of country, effective embargo, coalition planes and troops openly fighting it out with uniformed iraqi forces, both sides using tanks and artillery, sometimes in urban areas.

3: post-war: coalition (later iraqi govt) in charge of country, no sanctions, coalition troops and planes fighting various un-uniformed insurgents, non-state terror, crime, etc.

You seem to be making a claim about 1 versus 3, but surely the lancet study shows nothing of the kind, as it lumps together stages 2 and 3, which have very little in common.

All the talk of cluster sampling and error margins, and even misuse of the word civilian, pale behind this basic sleight of hand by which the large number of deaths from the boundary period between the two regimes are entirely assigned to one and not the other.

You could use a similar statistic trick to show cancer is better than non-cancer, if the surgery required to cure it was somewhat risky.

soru

Seixon, please buy yourself a stats textbook and look up “cluster sampling” before you waste any more of peoples’ time.

Your video is alleging that they found an effective method in faking news reports; I’m sure this is mostly a lie, but if it is true – that’s great, more power to them.

The Lancet figures imply 33,000 violent deaths excluding Falluja (violent deaths, not the 100,000 “excess” deaths that have gotten more attention). With Falluja, the Lancet figure is 189,000 violent deaths. The figure with Falluja is very imprecise due to the small sample size, so the Lancet authors focus on the non-Falluja results. The Lancet authors describe their estimates as “conservative,” meaning they are likely to be too small and underestimate the death toll.

The ILCS study says 24,000 violent deaths (over a slightly shorter time period). Since they have a much larger sample, they don’t exclude Falluja or any of the data. These are figures for all of Iraq, so they should be unbiased estimate rather than an underestimate or an overestimate.

So we have the Lancet figure, which ought to be an underestimate, actually turning out to be similar to the unbiased ILCS estimate. It seems that one of the two studies is mistaken. I think it’s the one with the much smaller sample and the much poorer methodology: the Lancet study.

You say: the ILCS didn’t get any [Fallujah-like samples] at all

I am just dumbfounded by this claim. The ILCS was a large random sample of all of Iraq, and there’s no reason to doubt that it sampled Falluja and other especially violent areas. The kindest response I can think of is: you are misinformed.

Thanks to the IBC, it is quite possible to do a systematic analysis of when the most fighting occurred in Iraq. I have done such analysis, and it shows that the periods covered by the ILCS and Lancet studies were about equally violent.

According to the IBC figures, your memory is betraying you. Summer 2004 was actually one of the most peaceful times in Iraq since the invasion (until August or September). The events in Falluja that you mention occurred in November 2004, and are not covered in either survey.

Your other big error is to forget that civilian deaths were highest during the invasion itself and soon after, a period covered by both surveys. According to the IBC, 74% of civilians killed by US forces in the first two years of the war were dead within a month of the invasion (by May 2003).

That’s the Lancet study in a nutshell ladies and gentleman.