Lancet Post number 135

Gilbert Burnham on the revelation that the British government's scientific advisors told them not to criticize the Lancet study:

We never had any doubt that the study would stand the test of time, it was designed with a lot of input from people who are leaders in their field and it was carried out in as robust a manner as we could, under the circumstances, so I was always certain that in the end the materials and the results would stand up to very close scrutiny.

Richard Horton on the British government's response to the Lancet study:

Would it welcome the Hopkins study as an important contribution to understanding the military threat to Iraqi civilians? Would it ask for urgent independent verification? Would it invite the Iraqi government to upgrade civilian security?

Of course, our government did none of these things. Tony Blair was advised to say: "the overriding message is that there are no accurate or reliable figures of deaths in Iraq".

His official spokesman went further and rejected the Hopkins report entirely. It was a shameful and cowardly dissembling by a Labour - yes, by a Labour - prime minister.

Mind you the Lancet denialists haven't given up. John Ray dismisses the opinions of Sir Roy Anderson, Chief scientific Advisor to the UK Ministry of Defence as "irrelevant" because they come from "A zoologist". Anderson's biography:

Professor Sir Roy Anderson was born in 1947 and received his BSc and PhD from Imperial College. Between 1971 and 1973 he was an IBM research fellow in the Department of Biomathematics at Oxford. He was Lecturer, Reader, Professor and Head of the Department of Biology at Imperial (1977-1993).

Professor Sir Roy Anderson was Head of the Department of Zoology at the University of Oxford from 1993 until 1998 where he was a Fellow of Merton College and is an Honorary Fellow of Linacre College; he was the founding Director of the Wellcome Trust Centre for the Epidemiology of Infectious Diseases at Oxford during this period. He was also Linacre Professor, University of Oxford from 1993 to 2000.

In October 2004 he joined the Ministry of Defence from his position as Head of Department and Professor of Infectious Diseases Epidemiology, Faculty of Medicine, at Imperial College, University of London. He retains a position as Professor of Infectious Diseases Epidemiology at Imperial College, London.

He was made a Fellow of the Royal Society in 1986, an Honorary Fellow of the Royal Statistical Society in 2004 and Foreign member of the Institute of Medicine, USA, National Academy of Sciences in 1999, Fellow of the Academy of Medical Sciences in 1999, Honorary Fellow of the Institute of Actuaries in 2000, Honorary Fellow of the Royal Statistical Society in 2001 and an Honorary Fellow of the Royal Society of Agriculture in 2002.

His publications include over 450 scientific papers in the areas of biomathematics and infectious diseases.

Instead of listening to Anderson with his "irrelevant" expertise in biomathematics and epidemiology, John Ray says we should listen to "statisticians experienced in the research method": Michael Spagat (an economist), Madelyn Hsaio-Rei Hicks (a psychiatrist) and Steven E Moore (a political consultant for Republicans).

Tags

More like this

Anjana Ahuja has written an extraordinarily one-sided article attacking the Lancet study. She drags out the same criticisms that were covered in the Nature story, but even though she cites the Nature piece, she carefully avoids mentioning the Lancet authors' replies, or the opinions of the…
A large group of public health experts has criticized the coalition for their continuing failure to count the civilian casualties in Iraq. In an editorial in the British Medical Journal Klim McPherson writes: Public access to reliable data on mortality is important. The policy being assessed---…
The Lancet authors have released the data to other researchers: Six months have passed since the publication of the study and we feel the time is now right to make the data set available to academic and other scientific groups whom we judge have the technical capacity to objectively analyze the…
I guess that the next time a new physics study comes out Science will ask epidemiologists what they think of it. You see, John Bohannon, the reporter for Science, decided that opinions from a couple of physicists and an economist were more important than getting comments from experts in…

Yep, they're denialists alright. I was considering adding something about them at my site but I've got to keep the topics within reason. There's just too many cranks to keep track of these days.

What's the emoticon for scrunching your eyes closed and refusing to listen?

I think that might be important to have as we wait for our delusionists to flood Tim's bandwidth with the e-equivalent of scrunching and widdle foot-stomping.

Best,

D

Lambert, throwing about statistics and pointing out who is qualified to judge them is an obvious troll for cranks. I'm sure this will bring in the climate audit crowd.

In this case it seems the denialists have won.

Man, that Roy guy has some serious CV. Spooky serious.

Qrazy, yes but don;t worry some guy called "Dave" will pop up shortly to say that he's found a misplaced apostrophe in a footnote on page 47 of a report Anderson wrote 30 years ago and he has evidence Anderson once voted Labour.

Then Dave will assert that he has equally impressive credentials but he won't provide his real name or what the credentials are.

Then Hans and Tim c. will rush out to embrace Dave for his courage and intellectual honesty.

By Ian Gould (not verified) on 30 Mar 2007 #permalink

Qrazy, yes but don;t worry some guy called "Dave" will pop up shortly to say that he's found a misplaced apostrophe in a footnote on page 47 of a report Anderson wrote 30 years ago and he has evidence Anderson once voted Labour.
Posted by: Ian Gould | March 30, 2007 08:23 PM

Is that really the pathetic extent of criticism of the Lancet Iraq surveys is it ?
Typos and attack the messenger based on their field ?

When you're finished talking about denialists (which I assume isn't about denial of bios) let us know if the figures in this Wikipedia entry aren't accurate.

That Lancet's figures would actually project 370,000 deaths for 2006, while actual deaths recorded are about 1/10th of that.

You're here talking about the politicising of research findings, yet not a peep about what would appear to be the worst smear seen so far.

I mean, either someone has completely misrepresented the most recent Lancet estimates as absurdly high or they just are absurdly high, right ?

There are no Lancet estimates without their surveys of deaths. These surveys reveal that 90% of the deaths are supported by death certificates.

Yet for 2006 the opposite is true. Lancet projects 370,000 deaths and Iraqi hospitals and morgues record only about 10% of this figure.

One could also point out that stats are one thing but actual, real, dead, bodies are another and that having 90% of your deceased civilians going missing makes the X-files look like PBS Frontline. But it's not like anyone has ever wanted to address that question to date, so you certainly won't see it now this figure is so implausbily high.

Yet someone criticising the Lancet figures as too high is "a denialist" are they ?

Really ? They're on par with those suggesting the holocaust, evolution and round-earth-theory are bullshit, are they ?

Would you honestly have no pause confirming that Lancet figures have this much variance from the official figures, on which they claim to be supported by, then claim that in 2006 (the worst year so far) that things are looking good because figures of actual people dying have dropped off so much that the Lancet projections are now off by so much.

Or would the more rational option be to figure out on your own that their pre-war death rate was clearly too low, therefor accounting for their simultaneous claims of 90% accuracy with official deaths and actually projecting a figure 10 times what those official figures are.

Either way, if you're talking about figures of the amount of dead bodies in Iraq and there aren't that many, isn't anything supporting the figures you are defending and absolutely no interest in corroborating those figures, you kind of a massive wanker to be calling critics of these figures "denialists".

Apart from the fact that it's a ridiculous slur against critics of a clearly flawed report, it would appear to be more applicable to you.

After all, is it geologists or fundamentalist christians who shy away from testing evidence relating to the fossil record ? Which side sticks with the original calculation for the age of the earth and refuses to question that figure regardless of how clearly that figure is contradicted by real-world physical evidence ?

Either you're pretending there's been a phenominal surge in interest in statistical methods and analysis or at some point you need to acknowledge that both promotion and criticism of these Lancet figures for deaths in Iraq are solely attributable to people caring how many people have actually died.

If you can do that and still maintain that the number of dead Iraqis is irrelevant then again, you're leading everyone in denial.

Kilo.

Let me in turn ask you a question: do you think that john Ray's description of Anderson as "a zoologist" is a fair description of his qualfiications?

If the case against the Lancet study is as strong as you think, why do you think its opponents feel it necessary to denigrate. misrepresent and trivialise the qualifications of a very senior and eminent scientist who was simply doing his job?

By Ian Gould (not verified) on 30 Mar 2007 #permalink

Let me in turn ask you a question: do you think that john Ray's description of Anderson as "a zoologist" is a fair description of his qualfiications?
Ian Gould | March 31, 2007 02:41 AM

I don't know who either of those people are, nor WTF the point of either answer would be. So let's say "no", that guy doesn't study animals.

In fact if it will avoid a similar post, let's say he's spent 80 years studying how many people die in Iraq and has varied phds in the same, remarkably specialised field.

So are the number of people here talking about "denialists" talking about people who are in denial about this random guy's bio or is that relevant to nothing I or they wrote about ?

why do you think its opponents feel it necessary to denigrate. misrepresent and trivialise the qualifications of a very senior and eminent scientist who was simply doing his job?

I don't. I don't think about them or what they write. It doesn't even interest me to the point where I can be arsed reading the linked articles or quoted crap on this page.

Dano:

What's the emoticon for scrunching your eyes closed and refusing to listen?

X |

On second thought maybe that is the emoticon for ferociously
determined troglodytism. I get them mixed up sometimes.

Kilo: "I don't know who either of those people are"

Then I suggest you read the first post and follow the links.

Have a good look around john Ray's site and you'll see why the term denialist is appropriate.

By Ian Gould (not verified) on 31 Mar 2007 #permalink

kilo:

I agree that the term "denialist" is not appropriate (certainly not as a general term) to describe all those who question the Lancet results.

Some of these people have legitimate questions about the Lancet study (and I'm not including the crowd who insinuate fraud was involved based on the "survey response rate" or meaningless speculation about survey "interview times").

Some of the people who are questioning the study results are also undoubtedly motivated by a concern for "how many people actually have died" in Iraq.

But the idea that they all are motivated by such a concern -- as you implied above with your statement "both promotion and criticism of these Lancet figures for deaths in Iraq are solely attributable to people caring how many people have actually died" -- is simply not credible.

When it comes to Lancet, there are certainly some (within the Bush admin, for example) who are motivated by something other than simply a concern about "how many people have actually died" in Iraq. Bush dismissed the results out of hand right after they had come out, for God's sake.

Kilo, I don't know what to make of the L2 number. I don't take the official statistics overly seriously--there are very strong incentives to lie. I've never heard of a government involved in a war that would freely admit how many civilians its own actions or the actions of its allies have killed--when they do acknowledge that such things happen they invariably give a figure lower than others give. And there are also incentives to lie even about the number of people killed by the insurgents, since the more they kill the less successful the war looks. Furthermore, I know of wars (Vietnam) where estimates of civilian dead vary by a factor of ten. And in this war it's not hard to find stories about the utter collapse of the Iraqi health care system and of course, stories about how dangerous it is for reporters to cover Iraq. So I would expect the official statistics to be much lower than the reality. Still, I do find it hard to believe that there has been a factor of ten coverup in this war. Hard, but not impossible.

Incidentally, a mistaken measure of the prewar mortality rate wouldn't solve the problem. L2 found 300 violent deaths from the invasion and occupation out of a sample of less than 13,000 people. Maybe they missed some prewar deaths--those postwar deaths are still there to be explained (and if they missed prewar deaths, maybe they missed a few postwar as well.)

To echo JB, your notion that every skeptic of L2 really cares about the truth is laughable. Many do care, but not all. And I don't think too many in either the government or the mainstream press care. As I say ad nauseaum, it really wouldn't be impossible for either the government or the media to do their own mortality survey on a larger scale and definitively confirm or refute L2's claims. Large scale polls continue to be done in Iraq. But none of those oh-so-concerned people in the government or the press can be bothered to do this, though it's been the obvious thing to do ever since the first Lancet paper came out and the controversy started back in 2004.

If critics with resources think that L2 is flawed and they really cared, they'd want to do it right. My suspicion is that the governments are afraid of the answer, probably fearing it will come out much higher than the official numbers whether or not it confirms L2. It might also reveal more civilian deaths caused specifically by coalition forces than has been reported, even if the true number turns out to be smaller than L2's. They obviously see no incentive to do an honest study and publish the results or they'd have done it. As for the mainstream press, the people with the resources to fund a poll (like the one they just did on Iraqi attitudes) institutionally it generally plays a lapdog role, only coming out in serious criticism on a given issue when it is politically safe to do so. That's how they behaved with the WMD issue.

By Donald Johnson (not verified) on 31 Mar 2007 #permalink

The "official" figures are at best a lower limit. There is no meaningful dispute about that. Many deaths do not get reported, the question is, how many? The criticisms seem to boil down to

1. The numbers are bigger than I want to believe, therefore they must be wrong. Somebody must be lying.

2. There are minor methodological concerns (e.g. "main street bias") that at worst could not conceivably account for more than a tiny part of the difference, therefore I can ignore the study altogether.

If the original study really was so incorrect, it seems to me that the US, UK, and Iraq governments would be rushing to fund a new independent survey with improved methodology. Yet oddly enough, the only people asking for such a study are the authors of the Lancet study. Now that we know what the statisticians consulted by the British government said about the Lancet studies methodology, I think we know why.

Kilo, I don't know what to make of the L2 number. I don't take the official statistics overly seriously--there are very strong incentives to lie.
Posted by: Donald Johnson | March 31, 2007 12:23 PM

For whom ? One would assume you're talking about the Iraqi and US governments as a whole. Even if you forget that the US def and intel committees also contain democrats with no such incentive and the Iraqi government contains ultra-partisan ethnic groups (each of which have been targetted by these killings), these aren't however the sources from which most information is obtained.

Brookings for instance has tracked the pissweak results of trained security forces. As the primary benchmark for progress, acknowledged or not, the number of capable Iraqi forces has been reported at embarrassingly low levels since inception and when pressed for examples of progress this number actually dropped from 3 to 1 battallions.
This is a highly subjective figure, apart from the fact you can lie about the readiness of troops freely without consequence, the Iraqi forces are also inflated by ghost employees. Yet you don't see this obvious incentive to lie, where there are no consequences of lying, taken up.
Dead people on the other hand are kinda more noticable and the people involved have no incentive to lie about it.

ie: It would be far easier and less likely to be immediately discovered as false that there were 1million trained Iraqi security forces (10 times actual) and that they were just otherwise engaged come parade time, than to claim there were 1/10 the amount of dead people and provide excuses for why people known to be dead aren't counted.

But this also isn't the primary source of info or cross-check for lies about Iraq. Journalists reporting leaked info from sources on all sides have broken stories about militia killings, death squads, aQ acting as defacto govt for whole towns, corruption, etc.

If you believe that in a climate where the number 1, 2, 3 and 4 stories about Iraq every day of the week for 4 years have been how many people died and not 1 journalist has yet discovered that only 1 in 10 deaths is being recorded then say this. If you think there's and incentive there to lie then say so. Simple.

Same for a Sunni hospital worker, like the ones you can find making documentaries about the awful state of the reconstruction efforts, who sees 20 Sunnis brought in with execution wounds every week just because they had the wrong surname on their ID at a militia checkpoint. Where's the incentive to lie for someone who knows their tribe is being ethnically cleansed, their time will come and it is being covered up in the meantime ?

How are you buying this with no evidence to support it ?
When the only reason to think this is to support a survey which clearly contradicts itself.

ps. What did you think of Colin Farrell's performance ? Moustache a bit much ?

But the idea that they all are motivated by such a concern -- as you implied above with your statement "both promotion and criticism of these Lancet figures for deaths in Iraq are solely attributable to people caring how many people have actually died" -- is simply not credible.Posted by: JB | March 31, 2007 10:44 AM

Yes it is. You're just confusing caring how many people are reported as having been killed as a result of your actions, with giving a shit about loss of life.

I'd refer you to Rumsfeld's outraged response to the Abu Ghraib story and how soldiers were allowed to use cameras as a clear example. Clearly he cared what was reported, just not the way other humans expected.

To echo JB, your notion that every skeptic of L2 really cares about the truth is laughable. Many do care, but not all. Posted by: Donald Johnson | March 31, 2007 12:23 PM
Why waste their breath then? Why go to the trouble of writing op-eds and unsolicited criticisms in the press ?
Clearly they care what is claimed as the number of deaths in Iraq, just not the way you think when you read the word "care".

What I'm saying is that the only reason Lancet is a story or an issue is that how many people have actually been killed by the war matters. Look at the title of this post. Now either proper sampling and extrapolation methods to arrive at a reasonable estimate are the issue to be addressed... or... the issue is how many people have actually died is.

I'm just wondering WhenTF this might enter the debate about the credibility of these Lancet estimates. Particularly when the criticisms cited are about spelling and education of participants instead of the fact that these figures are so wildly out of proportion to what is being recorded, yet claiming to correlate closely with that.

Someone trying to opt for an easy out by smearing another is clearly a pissweak arguement against the study. Are we thinking though that uncovering a bio is what's needed to avoid Lancet post number 235 or would actually validating this estimate of deaths with actual figures be a more useful objective. If so, join me in wondering why nobody has nor seems interested in even entertaining the idea.

"The criticisms seem to boil down to

1. The numbers are bigger than I want to believe, therefore they must be wrong. Somebody must be lying.
2.There are minor methodological concerns (e.g. "main street bias") that at worst could not conceivably account for more than a tiny part of the difference, therefore I can ignore the study altogether."

Ever notice how much of an overlap there is between global warming "skeptics" and Lancet "skeptics" and how similar their arguments are in both cases?

By Ian Gould (not verified) on 31 Mar 2007 #permalink

If critics with resources think that L2 is flawed and they really cared, they'd want to do it right. My suspicion is that the governments are afraid of the answer, probably fearing it will come out much higher than the official numbers whether or not it confirms L2. It might also reveal more civilian deaths caused specifically by coalition forces than has been reported, even if the true number turns out to be smaller than L2's.
Posted by: Donald Johnson | March 31, 2007 12:23 PM

You've answered your first sentence there with your next two as to why officials claiming the official figures are wrong wouldn't want to investigate the real figure.

As for the mainstream press, the people with the resources to fund a poll (like the one they just did on Iraqi attitudes) institutionally it generally plays a lapdog role, only coming out in serious criticism on a given issue when it is politically safe to do so. That's how they behaved with the WMD issue.

Well the WMD issue was being undercut at the time of the invasion and the govt being charged with outright lying 4 months in. You can't be suggesting that after a figure on 655,000 has already been reported without alternative (as opposed to challenge) that media feels it is politically unsafe to report on actual deaths in Iraq ? In 2007, when every story about Iraq for the past 4 years, regardless of topic, has included reports of deaths?

The "official" figures are at best a lower limit. There is no meaningful dispute about that. Many deaths do not get reported, the question is, how many?

If that's the real question you appear to not want it answered, given that you suggest only a couple of token and superficial criticisms which exist and then suggest that if the people who don't do bodycounts don't like the bodycount, they would do a bodycount. Circular logic and straw men. Awesome.

The criticisms seem to boil down to
1. The numbers are bigger than I want to believe, therefore they must be wrong. Somebody must be lying.

2. There are minor methodological concerns (e.g. "main street bias") that at worst could not conceivably account for more than a tiny part of the difference, therefore I can ignore the study altogether.
Posted by: trrll | March 31, 2007 06:49 PM

Really ? No criticism of merit or substance beyond "i don't like it" ? Only minor proceedural problems that wouldn't make much difference ?

As per that Wikipedia comparison this leads to this state of affairs for 2006 death estimates...

IBC - 24,500 violent

UN - 34,452 violent

Iraqi - 36,500 violent

Lancet - 339,500 violent

Lancet - 370,000 all causes

Now that Lancet violent figure is mine, being the 370k referred to earlier modified with info from Lancet's own Wikipedia entry where it states approx 601,000 of those 655,000 total deaths were violent. That's 91% of all estimated deaths in Iraq are violent.

So you'll note that with both the majority of the population and violence in the capital and immediately surrounding districts, along with the most reliable health services and press coverage, this figure of violent deaths estimated by Lancet is still around 10 times higher than those figures recorded by the press, and the health ministries which have have their records reviewed by NGOs and the press.

Are claims that someone is a zoologist really that lame when compared to a claim that 9 out of every 10 deaths, mostly occurring in Baghdad aren't recorded because health services outside of Baghdad are understaffed ?
That the press is mostly confined to Baghdad where most of these deaths are occurring and therefore can't get to remote areas where they are not ?

Where even the dead-of-night abductions, torture and execution killings have their bodies dumped in public for discovery. Were nuclear devices aren't yet in use by car-bombers and completely vaporised people aren't the overwhelming majority. Where even if it were carved-in-stone policy that all people killed by US troops were cremated and hidden on the spot it couldn't account for such and undercount.

Screw zoologist, I'll call the guy a fkn plumber if someone's willing to confirm they actually buy that crap. If they can tell me 90% is no longer a significant percentage. Or that ignoring a minority sample can account account for a 90% undercount in the majority. Or that claiming 90% agreement with figures you overstate by a factor of 10 isn't suspect, implausible and indicative that your estimates are fundamentally flawed.

WTF has that got to do with whether I like the figure of people killed in Iraq or sampling methods ? Anything ? Just a strawman you needed for an arguement against denialists. Hmmmm that sounds suspect.

Now you're not reading this blog unless you think that wildly divergent and unsupported claims deserve to be picked apart, exposed and ridiculed. Yet out of that list of figures there, one of them is supported and defended without reference to how accurately its estimates correlate with actual recorded data. Is that not a strange state of affairs for here ?

Am I missing something here ? Because it would appear an incredibly weak, hyperinflated and unsupported claim is being supported without reason.

Could ask why defend Lancet as established fact (ie "denialists" becomes appropriate) but more to the point, why not criticise it yourself ?
Otherwise it's gonna be a hard sell to claim that the term "denialist" is more applicable to other side.

"That's 91% of all estimated deaths in Iraq are violent."

No that's 91% of excess deaths above the pre-war baseline.

"So you'll note that with both the majority of the population and violence in the capital and immediately surrounding districts,"

The population of Baghdad is estimated at 5-6 million out of approximately 26 million.

The main areas of violence in "immediately surrounding areas" are located in Anbar, Diyala and Saladin provinces with a combined population of around 4 million.

So your basic facts are simply wrong.

By Ian Gould (not verified) on 31 Mar 2007 #permalink

There are some genuine questions about the attribution of violent deaths in the Lancet study.

An example of this is the high level of deaths attributed to US air strikes. (Some commentators have pointed out though that peopel on the receiving end of long distance weapons don't necessarily known whether they were attacked by aircraft or mortars.)

But the Lancet study measures all violent deaths. This includes not just violence that can be directly linked to the insurgency but the sky-high murder rate. (Much of which is probably ethnically motivated or the result of kidnapping and extortion by insurgents.)

The Iraqi Health Ministry on the other hand doesn't measure total violent deaths - it excludes all deaths that can't be definitively linked to the insurgency. so if a dozen guys turn up dead, handcuffeed, tortured and shot through the head, they aren't counted because they MIGHT not be the work of death squads.

The IBC figures include only deaths reported in the English language media - do you seriously think every single violent death in Iraq gets reported?

For that matter when IBC carries a report a day after an attack showing say 10 dead and 20 wounded, why do you assume that none of the wounded die subsequently?

By Ian Gould (not verified) on 31 Mar 2007 #permalink

"That's 91% of all estimated deaths in Iraq are violent."
No that's 91% of excess deaths above the pre-war baseline.

Yeah. Effect on what I wrote being nothing.

"So you'll note that with both the majority of the population and violence in the capital and immediately surrounding districts,"
The population of Baghdad is estimated at 5-6 million out of approximately 26 million.

Yeah that should have referred to the largest pop concentration rather than majority.

The main areas of violence in "immediately surrounding areas" are located in Anbar, Diyala and Saladin provinces with a combined population of around 4 million.

Brookings cites 28000 recorded deaths out of 44000 occurring in just Baghdad, without it's immediately surrounding provinces.

I'm quite happy using that as a basis for stating that the majority of violent deaths are occurring in Baghdad and it's immediately surrounding provinces, where health services and press coverage are also most competant.
I assume if you didn't also you'd say so.

Likewise for an absence of any direct statements expressing a belief in 300,000+ violent and uncounted deaths in 2006.

So your basic facts are simply wrong.
Posted by: Ian Gould | April 1, 2007 12:44 AM

Yet this doesn't appear to change what I've said about Lancet being obviously suspect in it's wildly disproportionate estimates. So great illustration of the "zoologist option" we were talking about earlier.

But the Lancet study measures all violent deaths.
...
The Iraqi Health Ministry on the other hand doesn't measure total violent deaths - it excludes all deaths that can't be definitively linked to the insurgency. so if a dozen guys turn up dead, handcuffeed, tortured and shot through the head, they aren't counted because they MIGHT not be the work of death squads.

Really? Are you sure ? Cause, you know, without those there is no such 100 per day total deaths per day during 2006.
I don't think it's going to fly if the suggestion is that the Iraqi official figures purposely undercount reprisal killings and overcount terrorist bombings to make up the difference for no apparent purpose.

See Brookings Iraq Index figures and compare these two:
p13 Jan-06 to present civ deaths from all violent causes.
p14 Civ deaths attributed to bombings
There aint no 100 deaths per day unless all killings are attributed as violent.

We're now in "main street bias" country, mentioned earlier. Where problems with minor variances (attribution of killer) in methodology are suggested to cause major variances (90% of deaths not being counted).

FYI the UN figure of violent deaths for 2006 of 34,000 is all violent deaths as per the cause of death recorded on death certificates not only by the Baghdad authorities but also provincial offices where these records play a role in death benefits, land claims, etc. Where the motication for this being counted and recorded is that of the family.

I'm not sure where the concept comes from that if somebody in the Iraqi government decides that these people were killed by neighbours, militias or US forces, how this would change the cause of death from a gunshot wound etc, but we'll assume it means nothing until it is explained.

You know, some might think that in a discussion about the number of violent deaths, that pretending whether known suspects were involved or not in a bunch of people winding up executed with gunshot wounds somehow changes the cause of death from violent to not, is a bit more misleading than the "your basic facts are wrong" examples you took issue with in my post.

The IBC figures include only deaths reported in the English language media - do you seriously think every single violent death in Iraq gets reported?
For that matter when IBC carries a report a day after an attack showing say 10 dead and 20 wounded, why do you assume that none of the wounded die subsequently?
Posted by: Ian Gould | April 1, 2007 12:55 AM

I don't assume anything about IBC figures. I think what they state about their methodology and it's inherent undercounting is and has always been quite clear.

If you wanted to think about IBC figures you could say that they form an absolute number for the lowest estimate, being a number of people killed in public and confirmed by two sources.

However death certificates are also a pretty good indicator of people being dead, and whether 2 news agencies covered them or not this doesn't change the deadness of the subject. Figures sourced from those sources have been crosschecked in a couple of different estimates, so IBC really isn't worth referencing.

Unless of course the idea is to take a known undercount and suggest it is evidence that there may be 10 times the amount of violent deaths that anyone has recorded dead bodies for. I would suggest in any other debate not one of you would be anything less than shamed to support such a proposition.

"Brookings cites 28000 recorded deaths out of 44000 occurring in just Baghdad, without it's immediately surrounding provinces.

I'm quite happy using that as a basis for stating that the majority of violent deaths are occurring in Baghdad and it's immediately surrounding provinces, where health services and press coverage are also most competant. I assume if you didn't also you'd say so."

So you maintain that "health services and press coverage are more competent" in Anbar (where significant population centres such as Fallujah and Ramadi have at various times been under effective insurgent control)rather than in, say, Irbil and the rest of Iraqi Kurdistan or such relatively peaceful southern provinces as Maysan and Muthanna?

By Ian Gould (not verified) on 31 Mar 2007 #permalink

As for "zoologist opinion", since you apparently still haven't bothered to check the initial link.

Anderson, the man in question, is a distinguished scientist who spent four years out of his entire career as a zoologist.

He didn't volunteer his opinion.

He was asked by his employer to take the Lancet report apart.

He wouldn't have done that by himself, he would have passed it to a group of statisticians and epidemiologists with instructions to find any possible flaw. (I know because this was essentially what I used to do as an economist working for the Queensland government.)

Going back to your employer and telling them you can't find any flaws is what Sir Humphrey Appleby used to refer to as "career-limiting behaviour". Before doing so there were probably two or three do-overs to confirm the lack of flaws.

That's why arguing that Anderson is a zoologist is not a valid argument.

By Ian Gould (not verified) on 01 Apr 2007 #permalink

So you maintain that "health services and press coverage are more competent" in Anbar (where significant population centres such as Fallujah and Ramadi have at various times been under effective insurgent control)rather than in, say, Irbil and the rest of Iraqi Kurdistan or such relatively peaceful southern provinces as Maysan and Muthanna?
Posted by: Ian Gould | April 1, 2007 05:58 AM

You quoted me pointing out that the majority of verified deaths have occurred in Baghdad.

I'll say Kim Jong fkn Il has been in control of health and press in the rest of Iraq if it will speed things up in terms of you not pissing about with irrelevant details and making even a single suggestion (we don't even have to start with a plausible one) for how Lancet estimates are not ridiculously high if they suggest a rate of violent deaths 10 times higher than figures based on the number of Iraqis recorded as killed.

Its telling that I even have to ask for this. If anyone actually believed this it would come up quite frequently in regular discussion about the Lancet estimates, given that it is the most striking feature of those estimates.

I really struggle to imagine another situation where this would pass without comment let alone scrutiny. A study suggesting that there were 10 times the amount of world oil reserves than any other study ? How about a study suggesting there was only 1/10th the amount of global warming recorded in other estimates ? Or closer to home, how about a study that suggested there were only 1/10th the amount of gun deaths than have been recored previously ?

Nah I think you'd see about the same amount of forthcoming explict statements of belief in those figures as you see for these Lancet ones. None. Pretty damn obvious to everyone why too.

As for "zoologist opinion", since you apparently still haven't bothered to check the initial link.

No, I haven't. Why would I bother ?
Spoil it for me. Does he say that one of the Lancet authors have admitted using incorrect methodology or that their numbers are by some other function bullshit ?

No ? Then that would really leave you with the same situation before anything linked in this post was reported wouldn't it ? The same study using the same methodology procuding the same figures it did last month.

I'll go to town reading the press release that says the report has been revised by its authors but until then all these criticisms and counter arguements count for shit.

That's why arguing that Anderson is a zoologist is not a valid argument.
Posted by: Ian Gould | April 1, 2007 06:21 AM

He could actually be a zoo specimin for all I care.
I referred to the "zoologist option" as where one takes a meaningless and irrelevant detail and tries to suggest that this undercuts an arguement which has no reliance on that detail. I referred to this so you'd remember it's already been pointed out here how lame that is and not continue with that option yourself.

Brookings cites 28000 recorded deaths out of 44000 occurring in just Baghdad, without it's immediately surrounding provinces.

I'm quite happy using that as a basis for stating that the majority of violent deaths are occurring in Baghdad and it's immediately surrounding provinces, where health services and press coverage are also most competant.

This sounds backwards to me. Since health services and press coverage in Baghdad are more competent, you'd expect a higher proportion of actual deaths to be reported there than anywhere else. That is, you'd expect under-reporting of deaths in the rest of the country, belying the claim that most deaths are occurring in Baghdad.

This is not an easy question to answer. There's good reason to think the official statistics don't capture the full death toll: a country that's gone through invasion, a radical regime change, insurgency and deadly sectarian conflict may have other priorities than rigorously counting all violent deaths and forwarding the information to the central government. Totaling casualties reported in the English-language media can only capture a lower bound: not all casualties are going to be captured in this way.

Polling samples have their own problems. When the misery is unevenly distributed, oversampling the worst or the best regions can badly distort the results. Eliminating this problem requires reasonably accurate regional population estimates. With hundreds of thousands of Iraqis either fleeing abroad or displaced within Iraq and a government struggling to perform its most basic functions, how accurate are such estimates likely to be?

With those caveats, two recent polls give us a clearer picture of conditions in Iraq: one carried out for ABC News, USA Today, the BBC and ARD German TV by D3 Systems, the other by ORB. Both sampled a large number of locations throughout Iraq.

In the ABC poll, 17% reported that a family member living in their household had been physically harmed by the "violence that is occurring in the country at this time". The ILCS survey indicated an average Iraqi household size of 6.6, so this implies a minimum average casualty rate of 2.6%. Iraq Body Count compiles English-language media reports of violence in Iraq, and says that 37% of reported civilian casualties to date are fatalities.

Media reports are probably more likely to underreport injuries than fatalities. On the other hand, some families may suffer more than one casualty. If casualties were distributed purely randomly this wouldn't affect the overall rate much, but they aren't. If one family member is targeted or living in a risky area, there is an increased chance that others are at risk as well. If the two unknowns roughly cancel each other out, that would imply violent deaths that were about 1% of the population: far lower than Burnham et al, but still a terrible human tragedy.

Baghdad is more deadly than the country as a whole: D3 said 77% reported they had had a friend or family member harmed in Baghdad, compared to 52% overall and 29% in Kurdistan. ORB reported that 51% of those surveyed in Baghdad had had a relative, friend or colleague murdered, compared to 38% in Iraq overall

42% told D3 they thought the country was in a civil war. Given a more nuanced range of options in the ORB study, 27% said the country was in a state of civil war, 22% that the country was close to civil war but not there yet.

D3 reported 42% thought they were better off now than before the war, 36% worse. In the ORB survey, 49% thought they were better off now, 26% that they were better off under the previous regime.

D3 said 48% thought the invasion was absolutely or somewhat right, 52% absolutely or somewhat wrong.

They overwhelmingly wanted the US to leave, but only 35% wanted the US to leave immediately

a single suggestion (we don't even have to start with a plausible one) for how Lancet estimates are not ridiculously high if they suggest a rate of violent deaths 10 times higher than figures based on the number of Iraqis recorded as killed.

Here are a few reasons why the official counts might be undercounts:

1. Inadequite staffing/training in Iraqi health system.
2. Poor communication between Iraqi health system and central counting officials.
3. Low priority of counting deaths in Iraqi health system.
4. Pressure for undercount by central counting officials.
5. Outright fraud by central counting/Iraqi health officials.
6. External pressure from U.S. government for an undercount.
7. External pressure from insurgent groups to not report certain deaths.
8. External pressure from a sectarian group to not report certain deaths.
9. Deaths not reported to the health system at all.

I'm sure there are some I'm missing. Could these account for a factor of ten undercount? I have no idea. But i'm not throwing out a well-regarded study just because I don't know.

Tim,

I am shocked and saddened to not have made your honor role of "Lancet denialists." Credit where credit is due, please. Could you please add me? Something like:

David Kane (statistician)

would do nicely.

And, for those Deltoid readers in Salt Lake City, I'll note that Les Roberts and I are both presenting (separately) at the Joint Statistical Meetings this summer.

Drop by and say Hi.

By David Kane (not verified) on 02 Apr 2007 #permalink

kilo replied to my post: "Yes it is. You're just confusing caring how many people are reported as having been killed as a result of your actions, with giving a shit about loss of life."

WTF?

"caring how many people are reported as having been killed"

is not what you said above (and what I responded to)

Above you said "both promotion and criticism of these Lancet figures for deaths in Iraq are solely attributable to people caring how many people have actually died"

There's a big difference.

8:35 am Monday, Kane? That's too bad. By Sunday evening, the conference goers will have figured out how to buy a drink in Utah, and, having spent the first few days of the conference not knowing how to buy a drink (Utah's liquor laws are sometimes confusing ...), they will have a few extra to make up for lost time ...

Kilo, as others have said (and I've said myself), I think the recent ABC poll strongly suggests a much higher death toll than the official figures, well into the hundreds of thousands of deaths. It's an open question to me whether the official figures are correct to within a factor of 2, 3, or 10. I agree it is hard to believe they are off by a factor of 10, but I don't know. Once you are willing to acknowledge (and I don't know if you are or not) that at least half the deaths are going unreported, where does one draw the upper limit of what is possible? I'd say millions of deaths is impossible, but hundreds of thousands isn't. I'd like to see more evidence before I'd commit myself to believing 600,000 violent deaths, but this wouldn't be the first war where official civilian death tolls were probably much lower than the reality.

As for the mainstream press, they have proven time and again that they don't do a very good job covering issues. There are individual exceptions, of course, but the civilian casualty issue isn't one of them. The first Lancet report on Iraqi civilian casualties came out in late 2004 and there have been a number of polls and surveys done of Iraqi attitudes since then, and yet despite all the controversy over the death tolls, no one in the US, UK or Iraqi government or in the mainstream press has commissioned a large survey to investigate the question of violent death rates. This is rather extraordinary if you take the self-image of the press as adversarial watchdog seriously-what could be more important than knowing how many people have actually died? The press is happy to pass on estimates and guestimates of extremely large death tolls attributable to our enemies (such as Saddam) without ever giving the basis for any of them, but when it comes to the deaths attributable to our own policies all of a sudden the evidentiary standard changes--only actual bodies (as reported by government officials anyway) are counted. This is absurd. If the press really cared about this issue, they would have commissioned their own mortality survey and they'd have done it on a larger scale and settled the issue once and for all. But they haven't thought to do this for the past two years, though I do recall reading about a survey done early in the occupation which concluded, without supplying details or confidence intervals or anything else, that Saddam had murdered some vast number of Baghdad residents. It was okay, I suppose, to ask that question and report the answer. The recent ABC survey that I and others have cited was concerned with Iraqi atttitudes and in the process of determining those it asked a question about whether a household member had been physically harmed, but this is no substitute for a mortality survey. So no, the press isn't seriously interested in the number of Iraqi dead, because they aren't using the one method they could employ to check the official numbers, a method they happily accept in other cases. And they don't have to accept the Lancet2 numbers if they don't want to--if they think the job can be done better, then why don't they do it?

By Donald Johnson (not verified) on 02 Apr 2007 #permalink

Will McLean wrote:

Baghdad is more deadly than the country as a whole: D3 said 77% reported they had had a friend or family member harmed in Baghdad, compared to 52% overall and 29% in Kurdistan. ORB reported that 51% of those surveyed in Baghdad had had a relative, friend or colleague murdered, compared to 38% in Iraq overall

This is a small but technically important point: you can't use these data to conclude that Baghdad is more deadly than the country as a whole unless you can also show that Baghdad residents have a circle of friends, relatives, and colleagues that is the same (or smaller) size as that of residents from elsewhere in the country.

David Kane said; "for those Deltoid readers in Salt Lake City, I'll note that Les Roberts and I are both presenting (separately) at the Joint Statistical Meetings this summer.

By any chance, is the following going to be your argument about the Lancet study at the conference?:

"The tone is unacceptable, the facts are shoddy, and the ideas are not endorsed by myself, the other authors on the sidebar, or the Harvard IQSS."

Oops, sorry, I was reading the wrong post on your website.

That was the reason Amy Perfors provided for removing your "A case for fraud?" post about Lancet.
Under "Thursday, March 15, 2007 SSS Removal"

Robert:

ILCS data was that Baghdad household size was about 10% less than the national average. Also, other measures of violence in the ABC and Orb polls also say Baghdad is more violent.

Not, however, twice as deadly as the national average, which supports the argument that offiical statistics and media reports are underreporting violence outside Baghdad.

Will, I was looking at your blog and I didn't realize that the ABC poll suggests that the Sunnis make up a larger fraction of the Iraqi population than previously thought. I don't know what to think about that (and have no right to an opinion, except a totally uninformed one, which has never slowed me down in the past).

I had read conservatives criticize the ABC poll because it had too many Sunnis, but since the poll results are broken down by sect (and the results for harmed households barely change if you use the more widely accepted breakdown of 60 percent Shiite, 20 percent Sunni and 20 percent Kurd) I didn't give it much thought. But it's interesting.

By Donald Johnson (not verified) on 03 Apr 2007 #permalink

Will:

I agree that household size in urban areas tends to be lower than elsewhere, but HH size is not the same thing as "relatives, friends, and colleagues." The ABC and ORB polls cannot be used to calculate rates of deadly violence in Baghdad or elsewhere.

Oh dear. Andrew Sullivan has linked to this post. Prepare for a loon infestation.

By Kevin Donoghue (not verified) on 04 Apr 2007 #permalink

Robert:

I think you need to distinguish between the published report of the poll results and polls themselves. The ABC poll did ask the narrower question of whether a member of the immediate household had been harmed: it just didn't present the Baghdad breakout for that number.

Will McLean wrote:

I think you need to distinguish between the published report of the poll results and polls themselves. The ABC poll did ask the narrower question of whether a member of the immediate household had been harmed: it just didn't present the Baghdad breakout for that number.

Will, I agree that Q35 of [this](http://abcnews.go.com/images/US/1033aIraqpoll.pdf) doesn't provide a Baghdad breakout. That's why it doesn't speak to whether Baghdad is more deadly, less deadly, or equally deadly. Am I overlooking something?

Robert:

The published report doesn't provide the Baghdad breakout for Q35. But the poll itself clearly collected the data, since the net of Q35 and Q37 is used to answer the question on page 5.

Will McLean wrote:

the question on page 5.

Do you mean this? "53 percent of Iraqis say a close friend or immediate family member has been hurt in the current violence. That ranges from three in 10 in the Kurdish provinces to, in Baghdad, nearly eight in 10."

Did the ABC poll ask how many close friends each respondent has?

A forthcoming talk by Riyadh Lafta seems worth a mention. I'll copy the announcement in case the link rots. Hopefully they will make a video or transcript available online. Note the bit about the refusal of a visa; have these guys no shame at all?

An Iraqi medical school professor will talk about the death count in Iraq after the 2003 invasion: causes, types of victims, categories of violence, and other health indicators.

Dr. Riyadh Lafta will be in North America to collaborate with University of Washington colleagues on a research project to document elevated levels of pediatric cancers in Basra, Iraq. The project was conceived as part of a sister university relationship between Basra Univ and the UW. The research project is supported by a grant from the Puget Sound Partners, a Bill & Melinda Gates Foundation initiative.

Dr. Riyadh Lafta, who teaches medicine at Baghdad's Al-Mustansiriya University College of Medicine, co-authored the October 2006 Lancet article that estimated more than 650,000 Iraqis have died as a result of the the American-led invasion in 2003.
Lafta will speak at a public gathering at Simon Fraser University's Wosk Centre (580 W Hastings in downtown Vancouver BC) on Friday, April 20, at 7 pm.

His talk will be video cast to the UW's Kane Hall at the same time with the opportunity for interactive audience participation.

Dr. Lafta will be a guest of Simon Fraser University, in part because the U.S. State Department would not issue him a visa to come to the United States.
The public is welcome at either location.

For more information:

Tim Takaro, MD, MPH, Simon Fraser Faculty of Health Sciences
ttakaro@sfu.ca
604-268-7186

Amy Hagopian, PhD, UW School of Public Health
hagopian@u.washington.edu
206-616-4989, 685-3676 or Ian Maki 206-543-6020

By Kevin Donoghue (not verified) on 07 Apr 2007 #permalink

Do you mean this? "53 percent of Iraqis say a close friend or immediate family member has been hurt in the current violence. That ranges from three in 10 in the Kurdish provinces to, in Baghdad, nearly eight in 10."

I'm sure there are some I'm missing. Could these account for a factor of ten undercount? I have no idea. But i'm not throwing out a well-regarded study just because I don't know.

Posted by: Abe G. | April 2, 2007 07:57 AM

How about throwing it out because of what you do know then ?

You know that at the same time it's estimates require such ridiculous suggestions for how a 90% undercount could occur, it tells you that the source data from which these estimates are drawn have a 90% correlation with these records.