Columbia Journalism Review on Lancet Study

Interesting that Lisa Guterman made no mention of the misrepresentation of the study's results by the study authors, or how the press -across the board- parroted this falsehood verbatim.

http://www.guardian.co.uk/Iraq/Story/0,2763,1338749,00.html
Guardian: "100,000 civilians dead"

http://www.nytimes.com/2004/10/29/international/europe/29casualties.html
NYT: "An estimated 100000 civilians have died in Iraq..."

http://www.washingtonpost.com/wp-dyn/articles/A7967-2004Oct28.html
"100000 Civilian Deaths Estimated in Iraq" -- WAPO

http://www.cnn.com/2004/WORLD/meast/10/29/iraq.deaths/
CNN: "100000 Civilian Deaths Estimated in Iraq."

etc etc etc. http://www.google.com/search?hl=en&lr=&newwindow=1&c2coff=1&q=%22100%2C…

Here's 18,000 more examples.

The "civilians" line was added by the Lancet editor; it's unfair to attribute it to the authors since it doesn't appear anywhere in the paper.

The "civilians" line was added by the Lancet editor; it's unfair to attribute it to the authors since it doesn't appear anywhere in the paper.

Fair enough, dsquared. My point really concerned the press coverage of the report. Tim and CJR seem awfully worked up that Kaplan is cited so frequently, but they don't seem to care all that much about the far more serious "100,000 civilians and probably more" meme regurgitated by just about every newspaper in the world. Or the misattribution of these deaths to "mainly" US airstrikes.

Tim's current bandwidth issues might have been predicted by his "flypaper for innumerates" post oh so long ago. I have read, literally, thousands of papers and have had almost a handful published with my name in them (two as primary author very soon). It is hard to write a paper. Usually you bring in someone from another department to help you. As it happens, one of my papers in review now needed a statistician to ensure my figgers were correct. Meeting with this person was humbling, to say the least - comparatively, I knew seemingly less than nothing about math (maths). So when Googlers chime in on boards like this and chatter about bias and undersampling, apart from being good for a laugh, we have to wonder about whether these electrons can be better utilized.D

The "civilians" line was added by the Lancet editor; it's unfair to attribute it to the authors since it doesn't appear anywhere in the paper.

Fair enough, dsquared. My point really concerned the press coverage of the report. Tim and CJR seem awfully worked up that Kaplan is cited so frequently, but they don't seem to care all that much about the far more serious "100,000 civilians and probably more" meme regurgitated by just about every newspaper in the world. Or the misattribution of these deaths to "mainly" US airstrikes.

Given that the authors concluded their study showed more than 98,000 excess deaths, I wonder why most media nowadays describe the study (when they report it at all) as indicating "up to" 100,000 being killed. Anyway, an example of the Lancet authors responding in writing to such reporting in The Independent (from an earlier post on Tim's blog):This letter from the authors of the study defending their findings was published December 12 in the UK newspaper The Independent on Sunday:Iraq toll: at least 100,000You reported ("Polish hostage held in iraq is released unharmed", 21 November) the Foreign Secretary's response to our study published in The Lancet of civilian deaths in Iraq. It is heartening that Jack Straw has addressed the topic in such detail. However, his response includes an apparent misreading of our results.Our study found that violence was widespread and up 58-fold after the invasion; that from 32 of the neighbourhoods we visited we estimated 98,000 excess deaths; and that from the sample of the most war-torn communities represented by 30 households in Fallujah more people had probably died than in all of the rest of the country combined.Fallujah is the only insight into those cities experiencing extreme violence (ie Ramadi, Tallafar, Fallujah, Najaf); all the others were passed over in our sample by random chance. If the Fallujah duster is representative, there were about 200,000 excess deaths above the 98,000.Perhaps Fallujah is so unique that it represents only Fallujah, implying that it represents only 50-70,000 additional deaths. There is a tiny chance that the neighborhood we visited in Fallujah was worse than the average experience, and only corresponds with a couple of tens of thousands of deaths. We also explain why, given study limitations, our estimate is likely to be low. Therefore, when taken in total, we concluded that the civilian death toll was at least around 100,000 and probably higher, not between 8,000 and l94,000 as Mr. Straw states. While far higher than the Iraq Ministry of health surveillance estimates, on 17 August the minister himself described surveillance in Iraq as geographically incomplete, insensitive and missing most health events.We, the occupying nations, should aspire to acknowledge the dignity of every life lost, and to monitor trends and causes of deaths to better serve the Iraqis, and in doing so, sooner end this deadly occupation.Les Roberts, Gilbert Burnham Centre for International Emergency, Disaster and Refugee Studies, Johns Hopkins Bloomberg School of Public Health, Baltimore; Richard Garfield School of Nursing, Columbia University, New York, USA[ends]I think this also suggests dsquared is being a bit harsh on Lancet editor Richard Horton: if the authors use the term "civilians" this freely when discussing their work, can Horton take all the blame for the many others who've applied it to the study?

Good spot and good points. Since I don't think anyone has suggested that there are more than 25K "combatants" in Iraq, the distinction is pretty silly anyway.

"My point really concerned the press coverage of the report."

Then why did you begin your remarks with the false claim that the authors had misrepresented the results?

Since I don't think anyone has suggested that there are more than 25K "combatants" in Iraq

Say what?! Fedayeen alone numbered at least 25k. Iraq's various insurgencies have been estimated at 200k. To say nothing of the hundreds of thousands in Iraq's regular army. This is a bizarre remark.

Then why did you begin your remarks with the false claim that the authors had misrepresented the results?

The authors of the Lancet editorial certainly did. JoT has shown that the authors of the study itself have also independently misrepresented its results in exactly the same manner.

"the authors of the study itself have also independently misrepresented its results in exactly the same manner."

Where? When? What exactly did they say (quotes in context, please)? Sources?

er -- Shirin scroll up a bit and you'll find the relevant passage I think: http://cgi.cse.unsw.edu.au/~lambert/cgi-bin/blog/science/LancetIraq/lan…

"we concluded that the civilian death toll was at least around 100,000 and probably higher"

Signed, Les Roberts, Richard Garfield, in the independent.

Fairly unambiguous hey?

Civilians or combatants? Does it really matter what label we give them? The study shows that there is a large probability that that the war in Iraq has resulted in at least 98,000 additional deaths. Regardless of the label they are still human beings who have died because of a war started voluntarily be western nations.

The study shows that there is a large probability that that the war in Iraq has resulted in at least 98,000 additional deaths.

No, the study shows that there is a large probability that the war in Iraq has resulted in at least 8,000 additional deaths. But I suspect many of us knew that already.

If you'd still prefer to sneak the Falluja results back into the study, and confuse the midpoint result for a baseline figure, I doubt dsquared or Tim will get on your case for it much.

Did you not read the comments by Lila Guterman at the start of this thread?

"With a quick call to a statistician, reporters would have found that the probability forms a bell curve-the likelihood is very small that the number of deaths fell at either extreme of the range. It was very likely to fall near the middle."

Using methods espoused and agreed to the world over by researchers there is a greater probability that the "true" figure for extra deaths in Iraq is closer to the midpoint (ie 98,000) rather than the extremes (ie 8000). These figures do not include the Falluja cluster. Read the abstract at the start of the article to know that this is the case.

Telluride: If we're gonna be picky about it, the figure of 98,000 and the lower CI of 8,000 refer to the deaths in the 97% of Iraq represented in the study by clusters other than Fallujah. So the low end of the range is 8,000 plus the Fallujah deaths.

Since most combatants will be healthy males, and for other obvious reasons, combatatants are more likely to comprise some proportion of the Lancet survey's 57,600 deaths from violence. But "combatant" doesn't just mean "insurgent" for a survey that covers the invasion as well as the occupation of Iraq, as has already been noted.The approx. 60K estimate above is derived from a total of 21 violent deaths recorded by Lancet ex-Falluja [table p4]. Just five of these violent deaths were recorded for the period from March 19th to May 1st 2003 [fig. p7], which could mean that soldiers comprised roughly a quarter of violent deaths, but could also mean they're not represented in the survey at all (we don't know that these five individuals were soldiers). And if all five were soldiers, how would we begin to estimate the number of civilians killed by violence during the invasion? These are the sort of fine-grained questions only a much larger survey or census can answer. I know I'm not alone in wanting to know how many Iraqi soldiers were killed during the invasion, not least because this might indicate how many orphans and widows were left in its wake.BTW while I've got your attention, dsquared, please note that it's Ramadi, not Ramallah, which is among the cities represented by Falluja according to Roberts et al. Ramallah is of course part of the other "occupied territories".

Who were these "biostatisticians"? And what did they actually say?

I've reread Kaplan's piece and the so called "debunking" thereof, and I find the latter far from convincing.

When Kaplan talks about a dart board, he's right. The wide confidence interval doesn't provide, well, much confidence. The use of clustering and the wide confidence interval are all just indications of how sensitive this study is, not just to random error, but also to other problems, as for example, all bombing deaths of women and children outside of Fallujah depend on just two interviews.

Kaplan's summary of the findings: "Assuming our model is accurate, the actual death toll might be 100,000, or it might be somewhere between 92,000 lower and 94,000 higher than that number." is much better than a claim that the study shows with at least "97.5%" probability that things have gotten "worse". The key phrase I'd emphasise here is "Assuming our model is accurate".

Kaplan is clearly "anti-war", and, in my humble opinion, gives the Iraqi Body Count far too much credit (they are far too generous in who they claim to have been killed by the coalition, suicide bombers are apparently sent on direct orders of Bush one might be led to believe), but he is honest in seeing the main concerns with the study, in particular, the wide confidence interval, sampling difficulties, the low estimate for the pre-liberation death rate (in summary, the information provided by the study does not add anything useful that we didn't know at least as well already).

Now as for the debunking:
1. The probability distribution within the interval. Kaplan didn't say that based on probability calculations for random sampling error all points in the interval would be equally likely. The point is largely irrelevant to Kaplan's argument.
2. Erring by not accepting the possibility of an "underestimation". Sorry, 50,000 to 70,000 killed through bombing in Fallujah and the like, which one would have to assume for serious "underestimation" are just not credible. We've got information outside the Lancet's numbers to decisively reject such a hypothesis.

I'm glad some of the people commenting here do not do statistics courses. They would fail.

Can you be a little more specific, Chorlto? As it stands your comment isn't very helpful.

Mike Harwood makes an interesting point. One of the deaths post liberation was attributed to the former regime. If we decided to reallocate that 1 death to the pre-war violent mortality, the factor of 58 cited by the authors would halve and pre-war violent death would double.

Most people commenting on the study, and that includes virtually everything written in this blog, engage in a politicised discussion of science of low scientific quality.

Show me one single detailed defense by a scientist of standing of the death rate estimates of the Lancet study, none has been produced yet to the best of my knowledge. I don't mean a quick phone call asking a scientist whether the authors appear to understand statistics, or a blog post on how malnutrition perfectly explains a near 3/4 drop in infant mortality in a few years, but a scientist of standing with relevant expertise explaining in detail, why say the CIA factbook estimates (which represent effectively the official US government position) are less credible than the ones of the Lancet study.

Hi JoT,

as happens we know that only 9 of the 21 ex Fallujah violent deaths were reported to have been blamed on the coalition.

5 of those were women and children dying in airstrikes (in two incidents). The authors describe the 3 shootings, so there is relatively little scope for soldiers to be included, maybe 1 (the man killed by an airstrike), but he might also be a civilian.

"One of the deaths post liberation (sic)was attributed to the former regime. If we decided to reallocate that 1 death to the pre-war violent mortality"

This makes no sense whatsoever. Why on earth would a study comparing mortality pre and post March 19, 2003 "reallocate" a death that occurred post-March 19, to pre-March 19, 2003 simply because it was attributed to the regime? That is just plain bad science.

This is just about the most absurd argument I have seen so far.

Reporters' unease about the wide range may have been a primary reason many didn't cover the study.

Nonesense - just another excuse. Does anyone doubt that a similar report (or even a much less rigorous one) about mortality in, say, North Korea or Iran would have been dicussed in the media for days, and then mentioned again at every opportunity?

In keeping with their standard policy, the mainstream media buried this report because it was unpleasant for them and for the government. The rest is just smoke.

Hi Pro bono,

yes, I believe reports on death rates for most countries in the world don't get much press.

Let's see, how many children under 5 do Unicef estimate die needlessly every year?

Do you know what the number is? How much press coverage it gets?

Hi Shirin,

why would one "reallocate"? For example, to see what the results would look like, if said single death had occurred a few days earlier?

It does grate to see a killing by Saddam's former regime in its dying days effectively blamed on the coalition (thereby raising the excess violent death toll by over 6,000, or more than 10%).

A lot of fun watching this from the sidelines - my non-expert opinion is the Lancet authors come off pretty well, except for this statement in the letter they supposedly wrote:

"Therefore, when taken in total, we concluded that the civilian death toll was at least around 100,000 and probably higher, not between 8,000 and l94,000 as Mr. Straw states."

I understand the reasons for concluding 100,000 is conservative, but there's a big leap between "conservative" and "minimum". Can anyone give me the reasoning?

Hi JoT,
My statement wasn't helpful nor meant to be. It was an exasperated sigh (aka dummy spit?). I suppose I was frustrated that we as humans tend to believe what we want to, when we want to. We will be swayed by our own beliefs and seldom will be swayed by arguments of others no matter how stringent and logical those arguments might be.

My position is that I have faith in statistical techniques to provide us with a robust and reliable model of reality. From this we can begin to understand the world around us. To whit I believe that close to 98000 extra individuals have died in Iraq as a direct result of the war. Itmight not be exactly 98000 but it is closer to this number than it is to 8000 (or for that matter 194000)

Heiko: I believe reports on death rates for most countries in the world don't get much press.

Sure, the media don't care about death rates in "most countries", but they do care about deaths in "rogue states". Remember, for example, the 1990s famine in North Korea? In case you forgot, you can just type North + Korea + famine into Google news and find out - in dozens of stories.

Millions dead, we are being told. How are these numbers arrived at? The media couldn't care less. North Korea kills people - that's a good patriotic story, that's all that matters. The US kills people - bad unpleasant story, move on, nothing to see here.

Well, Heiko, in my world when you are doing a scientific study you don't go around reallocating results just because it grates to see certain items on the "wrong" site of the ledger. In fact, a number of people died at the hands of Saddam's regime as a direct or indirect result of the invasion and occupation of the country. Some of them may have been killed before March 19,2003, in which case they belong on the pre-invasion side no matter how that may grate for some people or how it might affect the results. That's just how it is.

pro bono mathie,

You can bet that if the same study had shown either no increase in deaths or a very small one it would have gotten much more prominent play, at least in the U.S. media.

BTW while I've got your attention, dsquared, please note that it's Ramadi, not Ramallah

AAARGH! I haven't done it again have I? I have a total mental block on that one.

Serving members of the Iraqi Army are not likely to have been part of the households surveyed during the pre-invasion period, so I suspect that war casualties do not feature materially in the estimated deaths - unless I suppose the survey respondents disobeyed instructions and misreported them. But there's no real evidence to believe this happened.
Shirin: Or a negative increase in deaths! It seems hard to believe from where we sit now, but there was at least a theoretical possibility that the invasion could have made things better! (It did in the Kurdish North).

"the invasion could have made things better! (It did in the Kurdish North)."

I would not count on that either. According to the news I am getting from Kurdish, Turkmen, and Assyrian sources who either live there or have visited there in the last two years all is not as wonderful as you have been led to believe. Yeah, sure, the Americans are not bombing Kurdish cities into rubble and killing Kurds by the tens of thousands, but...

Pro Bono,

As I have said o this thread dozens of tmes, you've hit the nail on the proverbial head. Our media condition us to see the world through a one-way moral mirror. "They commit atrocities" whereas we "defend ourselves". They are "terrorists" or "insurgents" and we are "liberators" or "freedom fighters". These terms are rountinely used to describe any opponents or supporters of western foreign policy objectives. We certainly estimate numbers of "their victims" whereas we don't count "our victims" because, as Mark Curtis points out, they are unpeople in the eyes of the establishment. Foreign policy by the UK and US is based exclusively on "their use to us", and not by any scale of human rights. Right now we have "good Kurds" in Iraq and "bad Kurds" in Turkey, at least because both groups represent opposite positions with respect to western policy objectives. Thus, because Turkey is considered to be a partner in the "war on terror", their treatment of Kurds is irrelevant to western planners; by contrast, the Kurdish minority in Iraq are currently important, and hence are supported. State terror has no definition in western media circles. It goes on and on. It seems pretty clear to me that some of the contributors here who defend US aggression (except that they sanitize it by using other epithets) wish desperately to believe in the moral and ethical superiority of our so-called "democracies", and that our "elected" governments are universal supporters of these values and of human rights. A look at recent history blows this myth out of the water, but, like alcoholics who deny ther affliction, they want to believe that the "opposite of everything is true". Hence the constant nitpicking over the number of civilian dead in Iraq, as if the US/UK coalition really did everything to ensure that civilian casulaties were minimal, and that they were really interested in lberation and democracy.

Hi dsquared,

quick points:

a) If you believe the CIA factbook numbers, the invasion did indeed result in lower death rates overall. The CIA factbook numbers are the closest to an official US government position on the matter I've seen so far.

b) The Kurdish governorates were undersampled, 3 out of 33 clusters were located in the 3 Kurdish governorates. Consequently, the sample size is too low to draw the conclusion you have. The graphic illustration of the results in the Lancet study indicates a decline from 14 or so to 4 or thereabouts. A quick back of the envelope calculation shows that this is derived from something like 11 deaths before (responsible for about a quarter of pre-war deaths) and about 5 after.

Just for fun though, if the death rate in Kurdish areas had come down that much (which I don't believe is vaguely credible), the overall excess death should largely disappear merely by reallocating a few clusters, ie instead of giving the Sunni governorates of Anbar, Salahedin and Nineweh a total of 8 clusters, and the Kurdish governorates only 3, one were to allocate Anbar 0, Salahedin 1 and Nineweh 2 (and this allocation, while skewed towards Kurds, would seem at least as reasonable as the present one, and there was plenty of violence in Baghdad, which is allocated 7 out of 33 clusters, and responsible for 7 out 12 non-coalition violent deaths).

Heiko, you just pick at the tiniest of nits. The three Kurdish governates got three clusters instead of the four they would have got based on their population. Whoopee. I am not impressed by the way you try to make a mountain out of this molehill.

and this allocation, while skewed towards Kurds, would seem at least as reasonable as the present one

It might seem reasonable to someone who knew absolutely nothing about how the population of Iraq is distributed, but the study's authors were not in that category. Look the clusters were allocated to governorates (or pairs of governorates) on the basis of their population. Any other sampling scheme is going to give you results which don't have a prayer of representing anything.
Furthermore
Consequently, the sample size is too low to draw the conclusion you have
Why do you say that? Please, please, learn a thing or two about sampling theory before you go throwing around these sweeping statements about sample sizes. Three clusters of thirty households each, at an average six people per household, is a study covering 540 individuals. Since there is no reason to suppose that clustering effects would be material in the Kurdish regions, this looks like a decent size sample to me, particularly for such broad qualitative statements as "the observed sample death rate halved, so the underlying death rate is unlikely to have increased". What is your *specific* argument as to why this sample is too small?

Tameen also has a large Kurdish population and likewise got 0 clusters. Anbar, Nineweh and Salahedin got 8 all together.

Assuming the census figures are correct, and I've got my doubts there, the four Kurdish dominated provinces have a population of 3.553 million, and the 3 Sunni dominated ones a population of 4.7 million.

In the original distribution, the Kurdish governorates were given 5 clusters, and the Sunni dominated ones 6. However, Arbil's pairing was with Nineweh and Tameen's with Salahedin.

This is not nitpicking, it's potentially a material undersampling of an area known to have experienced particularly little coalition caused casualties, and a material oversampling of particularly troubled governorates.

On Suleimania's death rate numbers. 11 deaths and 5 deaths respectively is too low. Deaths can cluster for all sorts of reason. For example, among elderly friends, the death of one may precipitate several others in short order (similar lifestyle and age, and the traumatic effect of coping with grief). An accident may involve two people rather than 1, an outbreak of a nasty streak of influenza may strike down several elderly people at once. You know very well that the roughly 180 people in each cluster are connected to each other in various ways.

Heiko, you might gain a bit of credibility on this issue if you also mentioned the other governorates which had clusters removed; for example, Najaf governorate was paired with Karbala, removing any chance of a cluster in the city of Najaf and Qadisiyah was paired with Dhi Quar, moving the clusters from the governorate which contains the town of Samarra into a mainly Shi'ite province which did not see much fighting. If you're trying to turn one cluster in the Kurdish North into a massive influence on the whole study, you need to take these into account too.

Samarra?

http://en.wikipedia.org/wiki/Samarra

It's in Salahedin.

I don't think the moving around of clusters in the South is that significant. Most of the Kurdish North has seen essentially no fighting, as it wasn't invaded (and I believe that Tameen has seen much less than Salahedin or Nineweh).

Dhi Quar shows a very large increase in the death rate. It had four bombing deaths. Missan had 2 bombing deaths, and also shows an increase.

Your argument seems to boil down to a belief that the fighting in Najaf was so significant that this governorate would be likely to have a much higher death rate than the average of the Karbala, Missan, Wassit and Dhi Qar, and therefore this somehow makes up for the 2 clusters reassigned in the North.

I don't think that is all that credible.

Originally wrote this as a response to Brian S, before realising that I was answering the wrong question. So it's not a reply to Brian S. In fact, it's an argument suggesting that 100,000 is the right number to quote from the 8000-194,000 confidence interval. It's even "conservative", from a certain point of view. Since there seems to be some confusion over this, I'm (perhaps wrongly) assuming that this comment might still be useful.

Most (but not all) statisticians would say that being "conservative" means something like "when you are uncertain about what the truth is, you should assume the 'worst' thing is true". On that basis, you should choose your behaviour in such a way that your worst-case outcome isn't too bad. In the Lancet study, this could mean a few different things. If you are an American policy-maker, a conservative choice is to use the top of the confidence interval 194,000, because that's your worst-case outcome. If you're the author of a study claiming that "excess deaths" happened, then the most conservative figure is 8000, since that's the "worst-case" scenario for that claim (statistically, not morally).

However, it's more complex than this. When we draw a confidence interval around X, what we're doing is trying to find a band that X most probably lies in (actually, in the interests of intellectual honesty, I should say that this isn't exactly true, but it's close enough). The real difficulty is figuring out what we mean by "most probably" . A pure statistician likes to say 95%, or preferably 99% or better yet 99.9%. Because from a mathematical standpoint, that's "conservative". If your goal is to make sure you're not wrong, it's the right thing to do. In fact, the ideal answer is a 100% confidence interval, but the only way to get that is to allow a confidence interval that goes from minus infinity to plus infinity. Most of us (scientists included) find this to be unsatisfactory. We need to make real decisions, like "should I support or oppose the Iraq war?". So, if my goal is to answer that question, what should I mean by "conservative" and "most probably"?

My feeling is that there's no simple way to answer this, but I'll have a go anyway. Firstly, a "somewhere between minus infinite and plus infinity" confidence interval, though perfectly true, is absolutely useless for guiding action. If you want to perform any action in the world, you need to take some risk of being wrong. That's why no-one uses 100% intervals: they carry no information at all.

Ok. So, given that there's a war on, it is *not* conservative to throw up your hands and say "I don't know". You need to take a calculated risk given what you've seen.
So, how risky should you be? If you want to have an opinion about this matter, then you have to adopt some risk. However, to be honest, that's your problem. You pick your own confidence band. I won't presume to answer that for anyone else. Of course, the Lancet study authors have the same problem, so they can't do it for you either. Also, it's hard to start quoting confidence bands without making the snobby assumption that everyone in the audience has a stats degree.

What I could do is suggest (perhaps wrongly) that it's just as bad to (A) bomb people unnecessarily, as it is to (B) let a dictator murder innocent people when you can prevent it. So an overestimate of the number of excess deaths is just as bad as an underestimate. That's "conservative" for me to do (as someone "advising" you) because I don't know which of these things that *you* think is worse, or by how much (again in the interests of honesty, the "how much" bit is important for 100,000 to remain "conservative"). If that's true, then no matter how risky you are, no matter what probability of error you're willing to accept, X=100,000 will always be in the middle of your confidence band and also the value most likely to be true. Since *I* don't know where your boundaries are, and I don't know which type of error (A) or (B) you thing is worse, the only number that I can honestly quote to you is 100,000.

So one way to interpret "100,000 is conservative" is to suggest that the authors don't want to make any presumptions about what you believe about the relative evils of A and B, and nor to they want to guess what amount of risk you'll accept. Also, they don't want to assume that everyone who listens to them has a stats degree. Given that, the only honest thing they can do is to say that the most likely number is 100,000, but that there's a lot of noise in the data. What you do with that is up to you. What else could they really say?

Of course, they could've just meant that excluding Fallujah was conservative. I don't know. I didn't write the paper. However, it took me ages to write this without using the words "central limit theorem", so I'd love to know whether I should have bothered.

Thank you Brian, because the real value of the Lancet study is not as a post mortum for the Iraq war, but as a policy guide to the next.

Oops, that should have been thanks to Dan.

I have 2 questions for the stats gurus, if anyone can help out here. Since the 95% confidence interval refers to the 8,000-194,000 band, what confidence could be assigned to a somewhat narrower band surrounding 100,000, like 90-110k?

Also, what would the inclusion of Falluja data do to the confidence interval? Any ideas?

So much crap, so little time...
"To whit I believe that close to 98000 extra individuals have died in Iraq as a direct result of the war." Chorlto see, scientists ( as opposed to people who believe) are bounded by facts. If you truly believe that the number is close to 98,000 (e.g. +/- 10 k), then you have to accept that your estimate has a better than 50% chance of being flatly wrong. That's not science. Dan writes persuasively that we should ignore the confidence interval, because the 100k is in the middle. The reason he does so is because the study is so awful that the 95% confidence interval covers a 20-fold range. It just isn't adequate. the authors conclude in the independent that over 100k civilians died. They say this in the independent, because they would never have got that into a scientific paper. It is a revealing statement, because it shows that they have no confidence in their data. and when it comes to stats, the reason they excluded the Falluja data, is that it would greatly increase the variability of their data. If they include it, their 95% confidence intervals would include the possibility of 0 excess deaths. It is a post-hoc data manipulation, and it appears to be very biased. Heiko points out so many problems with the study, and he is right. It is just as bad as a whole slew of other poor epidemiology studies, and it got the attention it deserves. yours per

"To whit I believe that close to 98000 extra individuals have died in Iraq as a direct result of the war." Chorlto

Courtesy of a statistician on another blog, we have an answer to this question. A 16% confidence interval applies to the cited range. Chorlto has complete confidence in this 16% likely-event, which is why he always wins at roulette.

telluride, it seems you still do not understand the math.

Ah. Civilised debate. Normally when I comment somewhere before going to bed, I've been insulted when I wake up. I like this blog.

Eli: Thanks.

not creatively snipping: Just to clarify ('cos I'm a picky bastard), I'm not exactly saying that we should ignore the confidence interval, only that 98,000 (almost) always is your best "point estimate" irrespective of what confidence interval you choose. Knowing how noisy your data are is always a good thing. In this case, they're pretty noisy. The problem is that there's a standard formula for computing the confidence interval, namely 95% CI = mean +/- 1.96 standard deviations. So you can work backwards from their reported confidence interval and figure out that the standard deviation on here was about 47,500. For most people the standard deviation is a pretty abstract thing, but you can use it to calculate other things that you are more likely to be interested in. Which segues nicely into...

telluride: Yep. That's a 16% confidence interval all right. Well, when I did the calculations I got closer to 17%, but I'm happy to believe that the other folks were being more careful than me.

However, since I've got the software fired up, I thought I'd add a few other numbers that people might like. On the assumption that the authors' data are correct (on which I won't comment without reading their paper)...

Probability that the number of excess deaths is larger than...
98000. 50%
80000. 65%
60000. 79%
40000. 89%
20000. 95%
10000. 97%
0. 98%

That last number is a little different to the ones people usually quote when trying to determine whether 0 is inside the confidence interval. The reason is that I've changed the "intentions" behind the analysis, I'm allowing the upper bound on the confidence interval to be arbitrarily large, because I'm only trying to answer the "is X larger than..." question, which seems to be the one people are most concerned by.

A couple of other things. Although it's typical to report a confidence interval around the mean (98000 +/- 90000 in this case), we are often interested in other ones too. I don't know what other people are interested in, but here's a few that I think are helpful.

[0 to 20000] is a 3% CI
[20000 to 40000] is a 6% CI
[40000 to 60000] is a 10% CI
[60000 to 80000] is a 14% CI
[80000 to 100000] is a 16% CI
[100000 to 120000] is a 16% CI
[120000 to 140000] is a 13% CI
and they just get smaller from here on...

These things are additive (up to rounding error). So, [40000 to 140000] is a 70% CI. Of course, all this is based on several assumptions, the most important one being that the data are accurate. Which brings me back to...

not creatively snipping: The effect of reintroducing the Fallujah data would depend on what statistical methods they're using. To give a safe answer to that I'd need to read the paper, but there's a couple of things I can say. In one type of analysis (working with normal distributions), adding the Fallujah outlier would increase *both* the mean and the variability, so it may or may not extend the usual 95% CI over 0. In another type of analysis (working with binomial distributions), the effects are a bit different (for the stats folks: by this I mean that the sufficient statistics for the binomial are only the counts, while the normal has mean and variance as sufficient stats). In this case, my guess is that it would affect the mean more than the variability. I'd have to do the analysis myself to be sure, but it might be worth keeping in mind that the effects here aren't entirely predictable.

Hope this is useful. Again sorry for the long post.

Oops. That last paragraph should have been addressed to both not creatively snipping and telluride.

Dan
you wrote:
"Probability that the number of excess deaths is larger than...
98000. 50%
{SNIP}
20000. 95%
10000. 97%"
surely this must be different from what the authors have done, since the authors report 8,000 for their 95% confidence interval ?
yours
per

"not creatively snipping..." asks why Prob(N>10000)=97%

Answer: Prob(N<8000)=2.5%. You have to remember that also Prob(N>196000)=2.5%.

Actually, quite creatively snipping requires neglecting to point out that the authors reported 8000 to 196000 as their 95% confidence limit. Curiously the upper limit has fallen down the memory hole. Perhaps someone is pushing it?

However, allow me to toss some red meat to the cowering lions. The estimate of a mean of 98,000 in the distribution omits the Fallujah cluster. Yet it is certain that Fallujah and some similar cities were sites of major fighting in the past two years. Unless this was all happening on a Playstation it is both a real and statistical truth that the probability of the number of excess deaths being greater than 98000 exceeds 50%, though it is difficult to state by how much.

As to why the probability of exceeding the mean in a normally distribution is 50%, well, we learned that in fifth grade. Only a real snipper would have missed that class </snark>

Having turned the snark flag off, I would like to point out that the US armed forces in the 1980s and 1990s had studied and learned the lessons of Vietnam. Refusal to confront things like the Lancet study are quite worrying because it indicates an unwillingness to consider the results of actions on the political level. Basing policy on incomplete information is inevitable, basing it on false information is a disaster waiting to happen.

These probabilities themselves are based on assumptions, about the quality of the data, about Iraq's population etc...

Knowing that you can, and that's only possible through "judgement", work out your own confidence intervals, taking all the available information into account.

So my judgement is something like:

10% chance of lives saved
40% 0-50,000 excess deaths
35% 50,000 to 100,000 excess deaths
15% more than 100,000 excess deaths

(and I think the key uncertainty is non-violent mortality)

The IPCC does something very similar, when the data are far too complex to do an actual calculation that would have much meaning; eg they define "very likely" as an expert judgement of 90-99% likely.

not creatively snipping: Yep, I am doing something a bit different to the authors (Kevin's comment summarises the relationship nicely). Both analyses assume that we've got a classic bell-curve, with a mean at 98,000 and with a standard deviation of 47,500. The analysis they reported in their paper was a "neutral" one: they found the smallest range that covers 95% of the distribution. That's the 8000-196,000 range, giving us the usual 95% CI. The middle of this range is 98,000 because the bell curve is symmetric.

The problem is that the discussion we've been having on this site is about the lower bound (as Eli notes). To be charitable, I'll assume that this is because we're all agreed that 150,000 excess deaths is horrible, and so is 300,000, and that both numbers would lead to similar policy decisions in the future (i.e., "do something different next time"). So the policy-related question is now different to the original scientific one. We now need to ask what the chance is that the number is above or below 0 (or some other number that would lead to a "do the same thing next time" decision).

Statistically, we now don't care about the upper bound. So that's what I was doing in the first analysis. When I say that the probability that X>20,000 is 95%, this is the same as saying that [20,000 to infinity] is a 95% CI. Presumably the authors didn't report something like that because (a) it's not the convention and (b) it only makes sense to do this once you've started asking questions specific to the lower bound.

[Aside to stats people: frequentists out there, please don't start talking to me about one-tailed/two-tailed hypotheses, adjustments to alpha, and so forth. As a Bayesian, I really don't care. I'm just assuming that the reported sampling distribution is about the same as what the Bayesian posterior would look like. So all my numbers are merely descriptions of the posterior, not hypothesis tests (hence the words post hoc tests should be avoided at all costs). And yes, the assumption of posterior normality is exactly as plausible (or implausible) as the assumption that MLEs are normally distributed].

Heiko: Agreed on the principle that other information needs to be incorporated, no strong opinion on what that info is or what it implies about the updated CIs.

Hi Dan,

for me there are several factors that put the excess death number in context, firstly its breakdown into:
insurgents, young men getting their hands on a car for the first time dying in traffic accidents, civilians killed by coalition forces, changes in infant mortality ....

And secondly, future excess death. The invasion won't stop having effects after just a year and a half (or now two years), or have effects just in Iraq. Either for good or bad, the effects will likely run into the millions of excess dead, or respectively, lives saved, over the next few decades.

And thirdly, there are no truely "similar policy decisions" in the future. Other possible regime change actions will be very different.

In this threefold context, I find a statisticial test on the likelihood of greater than 0 (or another number) excess deaths of little utility in on its own informing future policy actions.

Sure, the media don't care about death rates in "most countries", but they do care about deaths in "rogue states". Remember, for example, the 1990s famine in North Korea? In case you forgot, you can just type North + Korea + famine into Google news and find out - in dozens of stories.

Sure, North and South Korea had equal amounts of famine, they just decided to only report on what happened north of the border.

Hi Heiko,
Ta for the comments. I'll follow your structure in replying.

1. Context. I agree. I just don't feel competent to make any strong statements about it, since I've stopped following the war in the papers.

2. Future Excess Deaths. Again, I agree, but with a few quibbles (like I said earlier, I'm a picky bastard). The stats that you can come up with using the Lancet data are, as you say, only relevant to the period up to Sept 2004. However, as humans (rather than statisticians) we're interested in what happens next. One possibility is that Iraq turns into the "beacon of freedom" that Mr Bush talks about. Another is that it decays into a hideous civil war. So when evaluating the value of the invasion, "what happens next" matters. But (to take the logic a little further), we also need to think about what would have happened anyway. Had we done nothing, the regime would have eventually fallen (all governments do eventually). Granted, it was probably years off (maybe decades: I'm not qualified to comment), but it's that alternative history that we should compare the war to.

3. Value of Stats. On this one I disagree (a bit). If you're saying "no two policy decisions are the same", I'd agree. But "similar"? I think that there are relationships between events that can be very useful when making decisions. Rather than pick a controversial event, I'll go back a couple of millenia. In my opinion, there are similarities (and differences) between the Peleponnesian wars (Athens v Sparta) and the Punic wars (Rome v Carthage). In both cases, one nation (Athens, Rome) was culturally (and, as it turned out, militarily) dominant over the other, with a much larger territory than the other. In both cases, the other power could plausibly have won the wars, and gone on to shape our subsequent history. Sparta and Carthage both had military advantages over their opponent (discipline & infantry for Sparta, naval prowess for Carthage), but never managed to use them as effectively as they might have. Had I been a Roman senator, I'd have studied Athenian history closely before starting a war or building an empire.

Bollocks. I'd only meant to preview, not post. The rest of part three was supposed to go something like this: Statistics is no more (or less) than an attempt to quantify this process of "studying the past". I suspect that future wars that America engages in will bear some similarities to the current one, though they won't be the same. I've no doubt that American policy-makers will look carefully at the statistics pertaining to this war before planning another one. Whether that's a good thing or a bad thing depends, of course, on your perspective. I'm only talking about the stats.

Let us not forget that the war was preceeded by a discussion about how only a minimum number of troops were needed to win, and a lack of discussion about how many were needed to occupy Iraq. The general officer who raised the latter issue (Shinseki) was put down and retired unceremoniusly. Even on the eve of the war, when Turkey refused to allow opening of a northern front, the US government did not wait until those troops could be usefully repositioned, but rather used the wishful excuse of trying to decapitate Iraq by bombing a location where Sadaam was rumored to be.

In this context the results of the Lancet survey are exceedingly important. It clearly shows the dangers of being able to win a war but not an occupation, which is why it is meeting with such a full throated renunciation based on wishful thinking.

The issue is not only the difference in costs between having a war or having Sadaam, but also the difference in costs between following a bunch of half cocked cowboy theories about war or a policy based on experience.

Let us not forget that the war was preceeded by a discussion about how only a minimum number of troops were needed to win, and a lack of discussion about how many were needed to occupy Iraq. The general officer who raised the latter issue (Shinseki) was put down and retired unceremoniusly. Even on the eve of the war, when Turkey refused to allow opening of a northern front, the US government did not wait until those troops could be usefully repositioned, but rather used the wishful excuse of trying to decapitate Iraq by bombing a location where Sadaam was rumored to be.

In this context the results of the Lancet survey are exceedingly important. It clearly shows the dangers of being able to win a war but not an occupation, which is why it is meeting with such a full throated renunciation based on wishful thinking.

The issue is not only the difference in costs between having a war or having Sadaam, but also the difference in costs between following a bunch of half cocked cowboy theories about war or a policy based on experience.

Eli,

my understanding is that the US didn't have (and doesn't have now) those extra troops available for the occupation. It would have required raising military manpower significantly.

Secondly, I don't think the problem is troop strength, but the kinds of troops available. Afghanistan has a tenth the troop strength for a similar size (population, territory) country as Iraq, and they are coping at least as well there.

I think that better trained troops (including speaking Arabic, crowd management, searching of civilians, checkpoint set-up etc...) would be far more important than more troops.

More troops would mean more targets for the resistance, and more opportunities for soldiers to shoot innocent civilians at extra checkpoints ...

To me it seems, the troop strength argument is purely motivated by partisan politicking (ie trying to make the Bush administration look bad, and by the way I am a German citizen living in the UK, and in the elections I am allowed to, I vote Labour, and if I was a US citizens, I'd vote Hilary in a race against Giuliani or McCain).

Dan,

on 1 and 2, complete agreement, including your "quibbles". On 3 I see your point (more on that in a moment).

Regarding the discussion of confidence intervals, you gave a differently worded explanation than mine, which I think is good, because rephrasing things and using different examples can help make them clearer. As happens, Donald still managed to misunderstand me, though I was saying the same thing as you were. I suppose I could have made it clearer which choice of estimating the variability of the data I actually supported.

Back to the "value of stats": I was talking about a very specific policy choice. Are we (the US + a coalition of the willing) going to pre-emptively (to exclude the possibility of counter-attacking after a nuclear first strike) invade and occupy another major souvereign country with an established leadership with a firm grip on power (this proviso added to exclude countries in open civil war such as Haiti or Liberia) with the aim of forced democratisation anytime in the near future (otherwise there is no urgency in getting the stats, conflicts in a few decades time are therefore excluded);

and would information on excess death in Iraq be likely to significantly assist in making that decision? (information on death rates is useful for a number of reasons, but how much of that usefulness has to do with informing the policy question just formulated)

The specific examples people come up with here are Syria, Iran and North Korea.

Of these, Iran and North Korea are militarily much stronger than Iraq was. A good case can be made that invading them would cost hundreds of thousands of lives (if not millions), no matter what the stats say about excess death in Iraq.

Syria's leadership could be overthrown like Iraq's with maybe similar military and humanitarian costs, if Syria was led by a Saddam Hussein like dictator, rather than Basher al Assad. But with him as the leader there, the justifications for such an action would be so weak, that even in the case of Syria, I reject the notion that we are in a desperately urgent need to find even more arguments speaking against a near term invasion without clear prior provocation of an extreme nature (annexation of Lebanon say, genocide of Kurds).

As said, everything you write there in 3 makes sense. I just added the above to explain what I was driving at specifically.

Afghanistan is irrelevant to the occupation (and the war) in Iraq, because there were numbers of Afghani troops on the ground who could (and did) participate in the campaign and the after campaign establishment of a provisional government (whether they should have been totally entrusted with pursuit of such a high value target to the US as Bin Laden is yet a further question).

There were no such native troops in Iraq EXCEPT in the Kurdish region, and that part of the campaign DID go well, which, I guess proves my point. However, it never was on that those troops could be used in the Sunni and Shia sections of Iraq because of ethnic rivalries. Chalabi's army was about as real and useful as the ceramic warriors of Xian, despite the fatal delusions of Wolfowitz, Rumsfield, Perle and Co.

Indeed, the lesson of the mini-wars of the 90s in former Yugoslavia was that advanced weapons wielded by a small number of highly trained, but not very large first world units, coupled to a relatively unsophisticated but numerous ground force could win wars and this WAS brilliantly exploited in Afghanistan. Anyone who had not figured this out by 1992 was either extremely stupid or extremely disingenuous.

The flip side, of course, was what do you do after you win. In the Balkens, in Afghanistan, in the Kurdish north of Iraq there were sufficient surrogates, but not in the south and the area around Baghdad. Indeed, as Al Gore observed, the time to have invaded Iraq was during the 1991 uprisings in the south, where armed native allies existed a plenty before being slaughtered from helicopters which could have easily been shot down by the US or its allies.

This brings me to my first point. There were two important questions about Iraq in the context of treats to the US, which should have been the primary drivers. The first was WHY Iraq. It has been shown conclusively post hoc that all of the justifications re US interests were BULL, both on the basis of what the US KNEW and what Iraq HAD. Many people, including me raised these questions before the US attack.

If the US had been significantly threatened the second question, HOW would have been less important to US policymakers and citizens. Of course, the war was sold by the policymakers misrepresenting the situation vis a vis WHY. That leaves us with assertions about what a nasty fellow Sadaam was, and he is.

But if WHY is not answered by a direct threat, the second question HOW, becomes paramont, and if there is a significant risk of making the situation worse (as the Lancet study shows it has), then the reason to go to war becomes weaker.

Yet, even in here there were possibilities that were not used. First, you are wrong to say that the 4th Infantry Division participated in the attack. They arrived afterwards from the Mediterranian and participated in the occupation, where, as you point out they were less effective. Had they been available in the initial attack troops would have been available to secure arms depots and stop the looting, two key conditions that aided the insurrectionists. Had the US allowed the UN inspection regime to go forward, in the face of Iraqi intransigence, Arabic speaking allies might have been persuaded to take part in a somewhat later campaign.

In short, other than the 2-3 week military campaign, Iraq has been a cluster fuck for the US from start to finish, and no walk in the park for the Iraqis either. So, let me in the same spirit as you show ask why you are dedicated to defending an action that has not only lessened the respect that the US is held in but materially weakened our ability to respond to the Bin Laden jihadists?

Hi Eli,

the Northern Alliance only held a very small fraction of Afghan territory and was also composed of ethnic minority groups, with the government to be overthrown actually having its support in the majority ethnic grouping (Pushtuns), while in Iraq, Sunni Arabs are a minority with just 20% of the population.

In the Balkans, there were and still are, a very large number of foreign troops, particularly per head of the population and for the size of the territory to be controlled.

As for the 4th infantry division, I didn't say anything about it. I don't think having had them available would have made as much difference as you seem to believe. Saddam had, to my knowledge, thoroughly distributed arms across Iraq, so no looting was necessary to ensure they'd be available everywhere, and I think extra troops would also have had to learn how to switch from combat to crowd control to fight looting and the breakdown of order.

So, why Iraq? To democratise the Mid East. Removing Saddam was much more defensible, and consequently feasible, than attacking any other Arab state, because of his past behaviour and the threat he still represented to the Kurds, Shiites and neighbours like Kuwait, Iran or Saudi Arabia.

With Iraq a democracy, the idea was (is) that the neighbouring Arab autocracies would more easily reform themselves without any need for further military actions.

The invasion was a pre-emptive action against terrorism, both to undercut the problems leading to terrorism (Mid East autocracies not doing enough for their people and allowing hate preachers to spread Salafi ideology) and to take out a potential state sponsor, who might eventually pass on the knowledge of how to make WMDs (or the weapons themselves, though less likely as more traceable) to terrorist groups.

In some ways it was a risky move, particularly with regard to the short-term propaganda hit against the US, and I only supported it, because I thought it would help ordinary Iraqis to a better life. So far, in my judgement, life is better on balance, but precariously so.

Heiko: "the Northern Alliance only held a very small fraction of Afghan territory and was also composed of ethnic minority groups, with the government to be overthrown actually having its support in the majority ethnic grouping (Pushtuns)"
True, but misleading.
Most of the land mass of Afghanistan is virtually uninhabitable high altitude desert. Conquering that part of the country is hardly a challenge - the Russians and Taliban both achieved that in a year or less.
The areas held by the Northern Alliance included a good chunk of the other part of Afghansitan, the relatively fertile valleys where the population is concentrated. The Northern Alliance controlled 10% of the land mass but a signficantly larger proportion of the population.
While the Pushtun are the majority in Afghansitan they're a bare majority 50-60%. The Tajik/Uzbek alliance that dominated the Northern Alliance made up around 30%. The balance of the population is primarily Hezaras who were also hostile to the Taliban.
In practice, broad ethnic identity is less important in Afghanistan than membership of or allegiance to a particular clan or tribe. The Taliban collapsed so rapidly because regional clan and tribe leaders who had previously supported them saw the writing on the wall - and an opportunity to resume their opium trading activities which the Taliban had banned.

I think Ian has addressed the issue Heiko raised about Afghanistan. <Tip of the hat>

Heiko's stuff about Iraq was so weak as to hardly require an answer. The press and the net is full of descriptions of looting in arms depots, which, I suppose is evidence that the Iraqi population was already armed to the teeth. The rest seems to posit that governments are supposed to make policy and take actions based on half-assed wishes. Nor is limiting the presence of extra troops when you don't already have enough something that gets you promoted outside of the French Army in WWI. I suppose the elan of the US troops was going to cow the Iraqi's not to come out on the street and loot. There were extra troops available. The 4th division was one of the best in the US Army, and it was not in position in time to take part in the invasion of Iraq. Reality 2 Elan 0.

Whatever.

All right, that makes sense.

I'd point out though that in Iraq Shiite clerics have a lot of influence, there are Shiite militias and among Sunnis there are also influential clans/tribes.

And an awful lot of Iraq is also desert. It's not that immediately obvious why a Shiite/Kurdish/some Sunnis coalition would need that much more support (in terms of troop numbers) than the Northern Alliance/Pushtun clan leader alliance in Afghanistan against the Taleban.

On the looting point, I remember descriptions of soldier blogs saying that Saddam had distributed weapons, explosives and ammunition carefully across Iraq in a manner that ensured there was no way it could be quickly secured.

I can't back that up with a quick google search though.

"an awful lot of Iraq is also desert."

And an awful lot of that desert has people living in it,or at least traveling through it at any given time. You cannot drive for more than a couple of hours or so on any road (or track, for that matter) in most parts of Iraq without seeing some kind of city, town, or village.

"It's not that immediately obvious why a Shiite/Kurdish/some Sunnis coalition would need that much more support (in terms of troop numbers) than the Northern Alliance/Pushtun clan leader alliance in Afghanistan against the Taleban."

It is perfectly obvious to those of us who understand the deep aversion on the part of Iraqi Arabs (and many Iraqi Kurds) to foreign dominance. The very idea that they would assist the Bush administration to invade, occupy and gain a foothold in Iraq.

"remember descriptions of soldier blogs saying that Saddam had distributed weapons, explosives and ammunition carefully across Iraq in a manner that ensured there was no way it could be quickly secured."

And of course they knew Saddam had done that because Saddam himself told them that, right? It can't be that they heard it from their commanding officers who heard it from their commanding officers, who were briefed on it by their commanding officers who were found it in the lists of talking points they received on a regular basis from the Pentagon.

Ooops - something happened to the end of my post before last:

The very idea that they would assist the Bush administration to invade, occupy and gain a foothold in Iraq is beyond unimaginable.

LOL well it doesn't take a genius to see rightists lining up on one side of the issue and leftists on the other.

Heiko, you are 100% correct. Iraq was one huge conventional ammo dump according to pretty much every soldier that's been there. As for why we didn't use the Shiite/Kurd militias: the biggest reason is probably that we were afraid they would wreck the country. In Afghanistan, there wasn't much left to wreck.

Shirin, again you're embarassing yourself. Saddam massacred Shiites and Kurds in the hundreds of thousands when they stood against him. The idea they would like some help is unimaginable? I guess you didn't see all those happy Shiites tearing down statues of Saddam alongside Marines, or the Kurds throwing flowers to our troops and waving US flags.

Tall Dave, how many Sunnis were there fighting us to a standstill in Najaf?

TallDave, I know this is going to come as a shock to you, but Saddam massacred more than his fair share of Sunni
Arabs as well.

But that is kind of beside the point. If you think the Shi`as are in love with George Bush and view him as their liberator, and if you think Shi`as are filled with deep and abiding gratitude to the U.S., and if you think the Shi`as are even remotely pro-American, then I have some prime ocean front property in Baghdad that I know you will find very attractive.

dsquared,

Anyone who knows Iraq's history, or who understands anything at all about Iraq, or about Shi`as in general and Iraqi Shi`as in particular knows with complete certainty that the only thing that has stopped a massive armed Shi`a uprising so far has been their obedience to their Taqleeds, led by Ayat Ullah Sistani. (That was also the sole reason for the high Shi`a turnout for the elections, as well - had he said don't vote, very few Shi`as would have voted.)

Shirin is correct.
Sistani is also elderly and unwell. It is vritually impossible for outsiders to predict who would succeed to his pre-eminent position in the event of his death - althoguh we can be pretty certain it won;t be Moqtada Al Sadr. He's too young, too unpopular in Najaf and the other pilgrimage centres and lacks religious stature. However there are other more senior figures who share many of his political views.

Ian Gould,

If it were not for his illustrious family background Muqtada Sadr would be just another religious extremist street thug dressed up as a Sayyid. He has no credible religious credentials, and the quality of his Arabic makes one wince. He is anything but a credit to his father, and he will never qualify to be an Ayat Ullah because he does not have the intellectual power or discipline to reach the level of knowledge required for that title.

I don't think we will see Sistani's like again for a very long time. He is, in addition to his religious credentials, a brilliant, wily and extremely disciplined politician who knows his own power, knows how to use it, and has a superb sense of timing.

dsquared,

That was a very very small percentage of Shia who were opposed by the vast majority, and they didn't "fight us to a standstill," they hid behind political agreementrs, died en masse whenever they attacked, and eventually fled the city, which was glad to see them go. How many Shia have joined the coalition forces in setting up a democratic Iraq? A lot more.

TallDave, I know this is going to come as a shock to you, but Saddam massacred more than his fair share of Sunni Arabs as well.
This may come as a shock to you, but the relative numbers are nowhere close. Of the three groups, Sunnis were by far the primary beneficiaries of Saddam's regime.

If you think the Shi`as are in love with George Bush and view him as their liberator, and if you think Shi`as are filled with deep and abiding gratitude to the U.S.
You mean like the 80% who supported the U.S. removal of Saddam in a Gallup poll conducted in March/April 2004? You really don't seem to know the first thing about Iraqi opinion.

Anyone who knows Iraq's history, or who understands anything at all about Iraq
That's clearly not you, from your earlier statements.

knows with complete certainty that the only thing that has stopped a massive armed Shi`a uprising so far has been their obedience to their Taqleeds, led by Ayat Ullah Sistani.
That's the silliest thing you've posted yet, which is saying a lot. Gee, you think maybe democracy has something to do with why there hasn't been a popular Shia uprising?

If you'd read a poll now and then you'd know most Shia are are very happy to see Saddam ousted by the U.S. and freedom and democracy on the way, and while they'd prefer not having foreign troops on their land, they regard it as a fairly small price to pay for their freedom from Saddam and hope for a better future.

Shirin: "the only thing that has stopped a massive armed Shi`a uprising so far has been their obedience to their Taqleeds, led by Ayat Ullah Sistani."

TallDave: That's the silliest thing you've posted yet"

Do you know what the word Taqleed means, TallDave? Do you understand its significance in Shi`i theology?

"How many Shia have joined the coalition forces in setting up a democratic Iraq?"

This is a perfect example of the kind of typical naive and superficial American way of looking at things that permeates the thinking even of most opponents of the Iraq adventure.

Very, very few Shi`as have "joined the coalition forces (sic) in setting up a democratic Iraq" (I will set aside for now the fact that the election was forced on the Bush administration despite its best efforts, and the process, which was designed by the occupying power for its benefit and not that of the Iraqi people, is far from democratic, and the election was not free by any of the accepted standards).

Most of the Shi`as who have "joined the coalition (sic) forces" in setting up whatever they are setting up left the country decades ago and look up residence - and citizenship - in other countries where they sat out the bad years in luxury, and then returned to Iraq on the backs of the conquerors.

The Shi`as I imagine you are referring to are not "joining the coalition (sic)" in anything. They are using the Bush administration to situate themselves in a position of political power, and they are doing so rather brilliantly so far. This is thanks virtually 100% to Sistani's combination of great power, wisdom in using it, and political genius.

"This may come as a shock to you, but the relative numbers are nowhere close. Of the three groups, Sunnis were by far the primary beneficiaries of Saddam's regime."

Typical uninformed, and inaccurate oversimplification. With Saddam it was quite simple, and quite simply not a sectarian issue. He gave benefits to those who supported him, punished with varying degrees of severity those who defied or opposed him, and did away with those whom he perceived as posing a threat to him. It was not a sectarian issue per se. Shi`as who supported him - and there were many - received the same rewards as Sunnis who supported him (and committed some of the most horrendously brutal atrocities on his behalf). Sunnis who defied, opposed him or posed a threat received the same treatment as Shi`as who did so. Some Sunni communities, such as Falluja, did not support the regime, and were punished accordingly.

The general Shi`a community was neglected and came under more punishment than the general Sunni community not for sectarian reasons, but because as a politically weak majority since Mesopotamia became part of the Ottoman empire, they have as a community historically been seen as a threat to those in power. This is an inevitable result of their opposition to the power structure, and the fact that they are a majority. The threat is amplified because of the Shi`a obedience to their Taqleeds. If a Taqleed issues an order that his followers defy or oppose a regime, they will do so. It is this opposition that led to the execution or exile of so many of the greatest Ayat Ullahs, such as Al Sadr and Al Hakim during Saddam's regime. Had they supported the regime they would have been rewarded instead of killed or exiled.

This is, by necessity, a greatly simplified explanation. To truly understand the politics involved here you really have to go back to the first century of Islam when the Shi`a split off from the rest of Islam.

"You mean like the 80% who supported the U.S. removal of Saddam in a Gallup poll conducted in March/April 2004?"

As you did with the AP story on the recent IRI poll, you are grossly misinterpreting the results here. It is a serious and dangerous error to mistake support for the removal of Saddam with support for the United States, the invasion, or the occupation. (And what poll are you referring to? This does result does not ring a bell at all.)

"You really don't seem to know the first thing about Iraqi opinion."

Oh, TallDave, I assure you that I know orders of magnitude more about Iraq, including Iraqi opinion than you could possibly learn in a lifetime.

"Anyone who knows Iraq's history, or who understands anything at all about Iraq That's clearly not you, from your earlier statements."

ma adri eish aqul laka ya Dave! Anta ta`raf kul shi `an al `iraq, wa ana a`raf maku shi. :o}

"That's the silliest thing you've posted yet, which is saying a lot.

Tab`an. :o}

"Gee, you think maybe democracy has something to do with why there hasn't been a popular Shia uprising?"

Democracy has nothing to do with anything that is going on in Iraq.

"If you'd read a poll now and then you'd know most Shia are are very happy to see Saddam ousted by the U.S."

The overwhelming majority of Iraqis of all kinds are happy to be rid of Saddam. That does not mean they are not extremely unhappy with what has replaced him.

"and freedom and democracy on the way..."

Wa kedha, wa kedha, wa kedha, wa kedha..............

Sirin,

Sigh... No, you clearly know very little except what you've been told to think, probably by leftist academes. Try some real facts instead.

"The only thing that has stopped a massive armed Shi`a uprising so far has been their obedience to their Taqleeds, led by Ayat Ullah Sistani?" Oh really? Here's a poll of Shia taken by Shia. This one was conducted by the school of political science at the Najaf University, polling 790 people between the ages 18-65 of both sexes and of different educational and socioeconomic backgrounds (and published in yesterday's edition of "Almendhar"):

"62% of those polled said they wanted Islam to be one of the sources of the constitution.

"38% wanted Islam to be the only source for the Iraqi constitution.

"49% support a federal government.

"50% support allowing those who boycotted the election to have input in writing the constitution.

"63% support the multi national forces staying in Iraq for the current time.

"85% expect the new transitional government to succeed in its goals.

"78% expect the new national assembly to successfully write a constitution by the dead line.

"1% said they expect civil war to break out."

Yeah, you can just feel the barely restrained rage simmering there.

All you have are totally unsupported assertions that are contradicted by all the real facts on the ground. It's nice you know a few local phrases, but in facts and perspective you are sorely lacking.

"Democracy has nothing to do with anything that is going on in Iraq."

Well, I guess you'll have to explain that to all those millions of Iraqi voters. Good luck, I'm sure your facility with the local language will prove invaluable in your quest.

As for me, I think I'm done trying to bring facts to the factless.

Ya Dave aT Tawil,

hal anta mu mafhum ila hessa?! yaba shuf. ma lazm astemi` ila mujami`in al janaH al mutaTarrif ay al janaH al yasari. HaSalat `ala al ma`rifa biT Tariqa aS Sa`ba ay min tajriba shakhSia li'anni `iraqiyya wa Hayati Hayat `iraqiyya. mafhum hessa? tab`an la li'nak ma trid tafham ay shi min al Haqiqa `an al `iraq. anta wa zumalak tafaDDalun taDhullun `umy wa Summ.

"Democracy has nothing to do with anything that is going on in Iraq."

Well, I guess you'll have to explain that to all those millions of Iraqi voters.

Nice to hear from you, Mr Mugabe.

Dear Commenter:

Equating voting with democracy is a common fallacy.

TallDave, is there any chance you could provide a link to your source for the poll you cited here? I would like to look at it in more detail.