Tim Blair

The Lancet article has been taken as gospel truth, but in fact it is a statistically insignificant pile of garbage.

The alleged 100,000 deaths could be as little as 8,000 (1 standard deviation) or as high as 198,000 (1 standard deviation), each admittedly with decreasing probability.

However, the most telling issue is the body count - 100,000 deaths over 550 days of the WoT in Iraq is 180 "innocent" deaths per day.

Not that I'm sure it is "innocent" deaths, either, as the survey was for "violent deaths" in the family, which could well include the family's 16 year old mujahdeen splodydope.

Where the left media gleefully reports 5 or 10 "innocent" Iraqui civilian deaths as a major media event, people, we are talking 180 bodies a day.

SHOW ME THE BODIES!!!

Strongly recommend the fisking of the Lancet article on the Chicagoboyz weblog. They have a whole series discussing the methodology and the findings of the article and leave no doubt at how really bad the Lancet article is. The pertinent postings are:

Links are in chronological order, earliest posts on top:

Bogus Lancet Study (Shannon Love)

Scientific Malpractice (Shannon Love)

The Lancet and the Iraqis (Ginny)

Judging Methodology (Shannon Love)

The Madness of Methods (Shannon Love)

Worse Than Nothing At All (Shannon Love)

Posted by Jonathan Gewirtz at 03:16 AM Permalink | Comments (1) | TrackBack (0)

Really interesting and worth looking into.

It's a little disturbing that some people WANT it to be true.

I thought it was supposed to be us warmongers who lusted for death. The left seems to be positively orgasmic over it. The bigger the number, the bigger the climax it seems.

That "right-of-centre Economist magazine" also endorsed John Kerry. Since a majority of voters voted for the other side, then presumably said "centre" is some distance to the left of the mean, median or center of the political spectrum.

Regarding political postion of the The Economist. I would agree with the Luther Blisset assesment .Its position mostly seems to be one of smug detachment.For example they might say that some peaple had doubts about the balance of BBC reporting of the Iraq war. Some of us would have described BBC reporting as relentless anti war propaganda.It is difficult to be right of centre in the gated community that is London media.

I cannot judge the statistics, but I will not trust any survey conducted by Iraqis. We know that security forces are infiltrated. We know there are always only women and children killed in American strikes because journalists who never leave their Baghdad hotel depend on Iraqi sources, often the same "minders" they had in the Saddam era. One interviewer or a few interviewed families with a grudge would throw the results off. This is speculation, but I think it´s reasonable to be very sceptical. Iraq is after all the country of the staged dead baby parade. What happened to the 500.000 dead children caused by sanctions?

The Lancet is not well-regarded in the trade these days, having published some poor-quality studies that were later discredited. It has an agenda. The MMR scandal is typical:

The Lancet MMR fiasco - summary of a scandal

[Management message: for God's sake, walter, you didn't put the goddamn url into the tag. If you can't use html properly, just PASTE THE FUCKING URL INTO THE COMMENT BOX AS IS SO PEOPLE CAN CUT AND PASTE IT INTO THEIR BROWSER ADDRESS FIELD AND HIT "GO" THEMSELVES. I swear if this keeps up I will turn off html entirely.]

It is worth noting that a problematic source like the Iraq Body Count website is good enough for the Economist, which used their numbers repeatedly.

Did The Lancet study remember to include the half dozen prisoners that Allawi allegedly off'ed, according to Paul McGeough's witnesses? It may be necessary to add 6i to the death count number.

This just in, German death rate increased after WWII started over the death rate during the same period before the war.

If +/- 1 Standard Deviation runs from 8,000 to 194,000 that means there is a 68% probability the actual number is between those two numbers. Scientific studies generally use a 95% confidence interval, if they use confidence intervals to analyze the data, which is almost 2 SDs (1.96). If the above limits are the interval for 1 SD, the 95% confidence interval includes ZERO. Generally when zero is includied in the confidence interval, one cannot say that the results of your sample differ from zero. In other words, their study showed that they cannot say that the rate of deaths in Iraq since March 2003 has had ANY increase over the rate before the campaign began.

When dead civilians are necessary to reproach US policy, dead civilians are supposed to be provided.
When possible, the US or its ally is faked into killing innocents, as, for example, when an armored column is fired upon from a school.
Colton and Palmer in their history of the western world, referred to the Balkan wars of independence in which various zealots would connive at getting the Turks to kill large numbers of whatever group the zealot was intent on freeing, "to dramatize their suffering in the eyes of Europe."
Should no actual dead civilians be available, lying is perfectly acceptable.
The most important terrain in any war since 1945 has always been the six inches between the ears of the American voter. Nothing is spared to conquer that territory.

Yes, the criticism of the Lancet study does seem to be politically motivated. For instance, in the first comment above, Kaboom demonstrates that he doesn't understand what "statistically insignificant" means and then links to Michael Fumento's grossly innumerate critique. I recommend Daniel Davies' comprehensive takedown of all the main attacks on the study.

If you want more, I have eleven posts on the study.

The medical treatment in the UK of an unfortunate Iraqi child who had lost limbs was a major news story.
If 100,000 dead, then perhaps 200,000 maimed.
The media wouldn't need to concentrate on the plight of 1 kid. They could be showing pictures of thousands of such victims. As they are not, I don't believe that they exist.

Always thought The Lancet was a British publication.

I remind everyone of the death count in Jenin.That too started out being based on "eyewitness" accounts.

The Lancet is indeed a British publication, but the study was conducted by Americans and Iraqis.

The problem with the cluster sampling counter-critique is that violent deaths in a war-torn country aren't like throwing rocks into a minefield, because we already know where the mines are. Military strikes are nonrandom, don't follow epidemiological patterns, and their locations are generally a matter of record. We know the Fallujah sample skews the results wildly because we know that Fallujah has been probably the biggest center of violence for both the coalition and the opposition. I am quite happy to impugn the integrity of the researchers, as well as the editors who let such a travesty be published, because even an undergraduate science student should have known the technique was inappropriate to answer the question.

I figured that if you used the Lancet study scenario to study Saddam's mass killings, you could figure that we found bodies in at least half the sites we looked at. Each site had an average of more than 600 bodies. We can now extrapolate to the entire country of Iraq, divide by the average mass grave size (appears to run about 100 yards by 20 yards), estimate that, per our example, half the country covers mass graves, multiply by the number of 100x20 yard plots that would make and then estimate about 7,000,000,000 killed by Saddam. Sure, that's 1,000,000,000 more than there are people in the whole world, but, what the hell. It makes as much sense as the Lancet article.

Tim:

Fallujah's data is excluded as an "outlier". Why?
Assuming the historical rate of 4:1 of serious wounds to deaths, if it had been included, it would show that every person in Fallujah had been killed or wounded at least once, many killed twice. (That's a simplification : many would be completely healthy, but many would have to be killed dozens, even hundreds of times).

I hope that you can agree that this is not the case.

So why was the Fallujah data discarded? Because such an absurd conclusion would tend to discredit the methodology?

The survey was based on more than just the word of the people interviewed. 80% of those asked were able to produce a death certificate to confirm the death.

Tim Lanbert,

Perhaps I missed something. The article I read said specifically that no death certificates were produced.

If that is in error, please help us out with a link to the facts.

Alan, I'm not sure that your 4:1 ratio applies to bombing. You exclude outliers because they give unrealistic results. This does not mean there is something wrong with the methodology, which was pretty standard random sampling.

TedM, are you talking about the Fumento article? Fumento says they didn't check death certificates, but they did. Here is a direct quote from the Lancet:

"We attempted to confirm two non-infant deaths per cluster, but in four of the 33 clusters no noninfant deaths were reported, and in some clusters interviewers confirmed deaths in more than two households. In 63 of 78 (81%) households where confirmations were attempted, respondents were able to produce the death certificate for the decedent. When households could not produce the death certificate, interviewers felt in all cases that the explanation offered was reasonable eg, the death had been very recent, the certificate was locked away and only the husband who was not home had the key."

The Lancet study reminds me of when I told someone that 50% of statisticians graduated in the bottom half of their class, and she responded "Wow, where do you get these figures?".

The notion of paying locals to go looking for violent death stats (without documentated evidence, apart from a rumoured 63 death certs) is the first problem. In order to find english translators you need to hire the only locals with such skills. They're overwhelmingly former baathist employees. Skewed data anyone?

Next problem is the way other data disagrees.
Examples;
1). "Civilian deaths in Baghdad, between May 1 and September 30, 2003, that Human Rights Watch says merit an investigation: 94" (http://www.tinkin.com/xmas/warnumbers.html)
If 180 civilians where dying EVERY DAY, mostly in air strikes, the baghdad stat would be a lot worse. And HRW would be saying something.
2). HRW claimes that civilian casualties are NOT largely caused by air delivered weapons, but by armour & artillery; "ground-delivered cluster munitions were a major cause of civilian casualties, while air-delivered cluster weapons caused a relatively small number of civilian casualties." (http://www.hrw.org/reports/2003/usa1203/4.htm)
So why does the Lancet blame air strikes?
3). Reporting on the initial phase of the war, HRW even said (same link): "Human Rights Watch’s month-long investigation in Iraq found that, in most cases, aerial bombardment resulted in minimal adverse effects to the civilian population..."
It should be recalled that air operations have dropped considerably since Baghdad was captured 18 months ago. So how are they killing 180 a day?
4). Other evidence shows air operations in Iraq have dropped considerably since the end of the initial ground war with the taking of Baghdad. For instance, all aircraft carriers where withdrawn from the region in may last year and stayed away for some time. Furthermore, most evidence shows air strikes where often more intense BEFORE the invasion (this Guardian article is an example; http://www.guardian.co.uk/september11/oneyearon/story/0,12361,787813,00.html). With the air power gone, how does the Lancet attribute so many casualties vaguely to 'air strikes'?

Mr Lambert, you aren't reading the evidence properly either. For instance, you posted:

"The survey was based on more than just the word of the people interviewed. 80% of those asked were able to produce a death certificate to confirm the death."

Then you posted later;
"In 63 of 78 (81%) households where confirmations were attempted, respondents were able to produce the death certificate for the decedent. "

63 out of 988 households in the survey isn't '80%' of the total. It's 81% of a select bunch.

I have some doubts about this evidence too. I'm not sure a lot of Iraqi hospitals produce death certificates. This isn't a western country we're talking about. A good exampls is this old CS Monitor article; "Nor are hospital records - kept in the heat of war under intense pressure on doctors and staff - necessarily accurate, some observers warn. That means they probably underestimate the real scale of civilian deaths, although at the same time they may have recorded some combatant casualties as civilian ones." (http://www.csmonitor.com/2003/0522/p01s02-woiq.html)

Wilbur, you might like to check what the story says again. It says that in Falluja the deaths were mainly from air strikes. You might be aware that reporting from inside Falluja has been a bit spotty in the past few months.

Wilbur, you do realise that it makes no sense to ask for a death certificate when there has not been a death? The "select" group is largely the households where there had been a death.

"Wilbur, you do realise that it makes no sense to ask for a death certificate when there has not been a death?"

Hmmm. You probably should have thought that line a bit more carefully before posting. You might want to read the links I provided too.

Anyway, 63 death certificates out of 100,000 deaths means documented proof for less than 1% of the total.

Also, observe; "Dr. Gilbert Burnham of the Johns Hopkins study team. He said the team had excluded deaths in Falluja in making their estimate, since that city was the site of unusually intense violence." http://www.iht.com/articles/2004/10/29/news/toll.html)

And furthermore; the same link states that the original study blames most of the violent deaths on air strikes.

Human Rights Watch have had people crawling all over Iraq nit-picking over everything, but they cannot produce stats anything like the Lancet's.

I should note that the study's spokesperson, Dr Gilbert Burnahm, is described as "a co-dDirector of the Center for International
Emergency, Disaster, and Refugee Studies at Johns Hopkins University."

Unlike the Human Rights Watch team, he isn't experinced in bomb damage assessment, visiting scenes of violent deaths, or, it seems, air delivered munitions.

The Lancet's technique is not too dissimilar to that used by David Irving, who claimed that the bombing of Dresden killed 250,000 people. Despite the fact that Dresden's fire chief only ever confirmed about 18,000 (and believed a few thousand more where missing).

Note that the Lancet study was also released in a rush in order to get it published before the US election. Stating it has a margin for error between 8,000 and 198,000 doesn't give me much confidence.

LabRat at November 15, 2004 10:24 AM validly draws a false conclusion from faulty premises:

Military strikes are nonrandom, don't follow epidemiological patterns, and their locations are generally a matter of record.

It is not quite true that US military strikes are non-random. But the blast and fragmentation patterns of 250kg bombs are not.

Rational refutation will only get you so far with know-not-so-much blowhards like Michael Fumento. From what I have read his main interest in this debate is to use his lofty perch in the Olympian Heights of Big Media to egomaniacly squash bloggers who might have the gall to call him on his mistakes.

Here's a sample, which might turn some of the blog-ears on this thread fire-truck red in rage, of His Fumentoness's dismissive attitude towards the blogosphere:

Earth to Inkstain and Lambert: Other than Inkstain caring what Lambert says and Lambert caring about what Inkstain says (perhaps), nobody cares what either of you says. Not only are you fully contained in the blogosphere, you're actually in a much tinier realm than that.
Meanwhile of the many places my piece on the Lancet trash appeared is today's Daily News, weekend population above 500,000. You attack not out of a sense that injustice has been done regarding the Lancet report, but out of jealousy. But if you cleaned up your act, you might just find that somebody somewhere, even with a circulation of ten, would occassionally print you. Alas, you will not. You are a lost cause.

What a prize jerk!

Fumento's pompous and arrogant attitude would be bad enough even if his anti-Lancet arguments had any merit. Whats worse is that Tim Lambert et al have used scientific methods to utterly demolished F.'s critique yet F. does not have the guts or honesty to fess up his blatant errors and, shall we say, "oversights". Way to go, Fuemento!

Bloggers of all stripes should always gang up on Big Mediators who try to monster them with their high-status credentials. These legacies are no guarantee of truth as the Lott, Rather and Bellesiles scandals show.

I have a highly unpaid job posting over at Catallaxy which has a readership in the high...tens. But thats not the point! Opposition to Fumento should be grounded in something that transcends partisan ideology: the solidarity of the blogologues.

Jack, that's true only if you are sampling deaths within the effective radius of a Mark 82 warhead, which is less than 300 or 500 meters, including shrapnel. Inside that radius, yes, random factors apply.

If you are sampling deaths across an entire nation, the pattern of bombing is important because the scale changes. The pattern is important because it dictates where the casualties are, and that pattern is decidely non-random, being the results of the actions of two competing parties. It only appears random because the parties are in conflict, and obviously don't coordinate their activities. (I am not trying to be sarcastic by stating the obvious, just being clear.)

That's one reason why sampling in this situation is questionable; statistical samples are supposed to be random, by definition.

A country with only two cities has a war. City "A" gets 100,000 tons of ordance dropped on it, and City "B" gets 1,000 tons. A "random sample" is going to show more deaths in "A" than in "B".

And the researchers should slap their foreheads and shout "D'oh!!!"

As I recall my stats (years ago), a random sample study should include all parts of the nation, including those that are not engaged in combat. The response might be zero, or it might not be, if people go to fight in other parts of the country, or are killed while traveling (e.g., business). But the sampling is nationwide, not focused on combat areas.

In reality, the only way to accurately determine casualties in a war is to count bodies. And that is highly questionable in a nation with poor record record keeping (e.g., Iraq).

No, the count won't be complete. Even today, casualties from WWII and WWI are estimates simply because of the uncertainies of the sources, or the extreme devastation of weapons (excluding nuclear weapons, by the way -- note the comment above about Dresden).

That's why I find this whole study silly. Random sampling of a population whose belligerency can't be determined, focused on selected parts of a nation with non-existent record systems, during a war where even portable direct fire weapons can vaporize a human body? Please! It'll be years, if ever, this can accurately determined.

Anything else is, at best, an educated guess. Therefore, assumptions control. That's why I view this Lancet study as a joke.

Whoops! One clarification -- portables don't "vaporize" human bodies, but the end results is close enough. How do you determine how many people where in the house after tossing in grenades, it burns down, and gets crunched by tanks? Especially if no one knows who was in the house in the first place?

Let's be realistic here, please. God knows Lancet ain't.

The main source of confusion over the Lancet study is that it includes nation-wide increases in civilian mortality, not just theartre-wide instances of military casualites. People focusing on headline casualty figures are going to miss the latter issue.

Sitting below the ice-berg tip of military KIA/WIA's there would still be a huge submerged base of civilian mortality. This is caused by the war-induced breakdown in Iraq's institutional, infrastructural and industrial systems.

So even if US military authorities are accurate in their targetting there would be a world of civil pain to account for. Unfortunately, US military agenciesare resorting to the use of heavy weapons (aerial bombs and artillery) in urban areas which must invariably hit insurgents and civilians alike. This is partly to replace man-power with fire-power and partly to drive a wedge between Suuni rejectionists and jihadist incorrigibles.

In most 20th C guerilla wars the civil-to-military casualty ratios can climb as high as ten to one. Which makes Lancet's ~100 k civilian figure look pretty plausible.

I would submit that these high civilian casualty figures are one major reason for Iraqi popular hostility towards the US occupying forces. The other is the fact that the Suunis will never forgive the US for putting them under the thumb of the Shiites.

OK, Jack.....

...war-induced breakdown in Iraq's institutional, infrastructural and industrial systems.

If you said war-accelerated, you would have a minor point. The infrastructure in Iraq was in terrible shape after 30 years of neglect. I've followed a lot of the reconstruction work there (professionally -- it's part of my job). Add in the fact that the infrastructure was used to control the population, and the war impact is minor (except that the terrorists try to blow up the reconstruction efforts). Only parts of a couple cities had regular power, water, and sewage. Coalition efforts have been improving the infrastructure. Check here and here.

Unfortunately, US military agenciesare resorting to the use of heavy weapons (aerial bombs and artillery).

Even the commenters in this thread have noted that aerial delivered warheads are highly accurate. Remember precision guided munitions, Jack? Accurate to within a few feet? Try Google. Add in the targeting process, and collateral casualties are minimized. Not eliminated, but minimized.

Artillery....it's clear that you have zero experience with indirect fire weapons. Artillery is observed fire. You see a target, and you blast it, but it is controlled, almost as accurately as aerial munitions. Civilians were not targets. Enemy positions were -- and those were mostly known to within a few feet, thanks to satellites and airborne surveillance. Again, collateral casualties were minimized, not maximized.

In Fallujah, I doubt that artillery is being used much. Mortars and direct fire tubes (tanks), yes. Arty is of limited use in urban conflict, especially when your troops are also in the urban area. Fire control must be very precise, or you have "friendly fire".

In short, weapons technology reduces (but does not eliminate) collateral casualties. Operational planning and rules of engage reduce collateral casualties even more. Read this.

In most 20th C guerilla wars the civil-to-military casualty ratios can climb as high as ten to one.

First, this is the 21st Century. Second, this is not a civil war. Third, that's pretty damned high ratio. Where did you get that from?

Which makes Lancet's ~100 k civilian figure look pretty plausible.

No, it doesn't. Your rational does not add up.

To be fair to The Economist, they do not "praise the study unreservedly".

Quote:

The study can be both lauded and criticised for the fact that it takes into account a general rise in deaths, and not just that directly caused by violence. Of the increase in deaths (omitting Fallujah) reported by the study, roughly 60% is due directly to violence, while the rest is due to a slight increase in accidents, disease and infant mortality. However, these numbers should be taken with a grain of salt because the more detailed the data—on causes of death, for instance, rather than death as a whole—the less statistical significance can be ascribed to them.

The study is not perfect. But then it does not claim to be. The way forward is to duplicate the Lancet study independently, and at a larger scale.

[Management message: for God's sake, walter, you didn't put the goddamn url into the tag. If you can't use html properly, just PASTE THE FUCKING URL INTO THE COMMENT BOX AS IS SO PEOPLE CAN CUT AND PASTE IT INTO THEIR BROWSER ADDRESS FIELD AND HIT "GO" THEMSELVES. I swear if this keeps up I will turn off html entirely.

Ouch.

http://briandeer.com/mmr/lancet-summary.htm

JeffS, it would be better if you criticized the actual Lancet study rather than the imaginary one you invented. They don't select clusters but randomly sample them. They give details of how they sample. If you think there is something wrong with their sampling procedure, share it with us.

That's better, walter.

Tim, perhaps I misread the study. My understanding is that the researchers focused on the areas of Iraq with active terrorist operations, rather than random samples across the nation, including relatively peaceful areas.

If I read that correctly (and I'd best go back to do so), the data are going to be skewed towards the violence. Perhaps I'm wrong.

I'm also amazed that the researchers aren't questioning the credentials (for lack of a better term) of their samplers, as
walter notes.

A valid statistical analysis requires representative sampling techniques -- look at the problems with the exit polling in the US presidential election.

This is above and beyond the standard deviation of error produced by the analysis. An SD of 190,000 casualties? That alone cries for a peer review before publication.

The researchers randomly sampled the entire country. The article gives lot's of details.

Walter seems to think, based on the notion that only Baathists speak English that the interviewers were Baathists. He has no evidence to support his theory. The interviewers were hired and trained by the two Iraqi authors and five of the six were medical doctors.

The study was peer reviwed before publication. It did not have an SD of 190,000 casualties. The SD was about 45,000 deaths.

That would be aproximately 100,000 +/- 92,000.

That is a large MOE especially for a 68% confidence level. It also means (correct me if I'm wrong) that there is a 16% chance it could be less than 8,000. (about 1 in 6).

The Real JeffS at November 15, 2004 02:48 PM makes some valid points but not decisive ones.

I acknowledge that a decade of (erroneousy imposed) UN sanctions did most of the damage to Iraq civil society and material structures. However the US's shock & awe military violence has clearly pushed many Iraqis over the tipping point into the mortally-risked range of conditions.

I acknowledge that US operational planning and rules of engagement oblige the use of guided munitions in urban warfare. I also insist that whilst munitions can be guided, shrapnel and blast effects cannot. The US's hi-tech bombs and shells are still immensely powerful, more so than in the old days of WW II. The use of these heavy weapons in an urban environment is in violation of international law and conventions that constrain the military from causing predictable harm to civilians.

This less-than-discriminating use of firepower does not help the US or the cause of Iraqi freedom. It may help Bush's political prospects.

The source for the 20th C's ten-to-one civil-to-martial casualty ratio is a factoid that is floating around the internet that I have soureced to a Dr Spiegel a senior associate at the Johns Hopkins University in Baltimore:

Spiegel reported that the proportion of all civilian war casualties has increased from approximately 14 per cent in the First World War to 67 per cent in the Second World War, and to 90 per cent in the 1990s.

Finally we should, in view of the irregular procedures which characterised the US's launch of a preeventive war, take into account the numbers of Iraqi regular and irregular soldiers who took up arms against the US and have been killed in this conflict. They are lawful combatants and have a right to be accounted for.

Bob Woodward, the Court Historian of the Bush admin, interviewed Secretary of Defense Rumsfeld and discussed Gen. Tommy Franks estimates of the total Iraqi army death toll suffered up until the fall of Baghdad (April 9 2003):

Woodward: April 9th the day the Saddam statue came down in Baghdad. At the briefing General Franks to the NSC estimated that there had been 30,000 casualties that’s the number he gave. You have any reaction at the time?
Rumsfeld: I don’t know what you’re saying, 30,000 casualties?
Woodward: 30,000 Iraqi soldiers killed – I’m sorry, it was killed not just casualties.
Rumsfeld: I don’t remember it,

There would have been a lot more Iraqi battle bodies counted in the past 18 months. So the total Iraqi civil and military casualites arising from this confict, including both lineal and collateral victims, and taking into account the high social to martial ratios, is quite plausibly close to 100,000.

I was rubbing my eyes in disbelief for a while too, but the numbers make sense to me now.

M. Simon, 8,000-194,000 is a 95% confidence interval. A 67% confidence interval would be about 50,000-150,000. If the war, on net, saved lives there is a less than 2.5% chance that you would have seen numbers like those obtained.

I've spent a couple of hours reading the original Lancet document. I still can't believe people are taking it seriously.

I don't even know where to start with the problems in it, but I guess i could begin with this quote from the document itself;

"the deaths reported by the remaining families might represent a disproportionate number of deaths from the larger community that used to live in the area, leading the interview data to overestimate mortality."

And this one:

"The cluster survey methodology we used may have, by chance, missedsmall areas where a disproportionate number of deaths occurred, or conversely, selected a neighbourhood that was so severely affected by the war that it represents virtually none of the population and thus has skewed the mortality estimate too high."

and this;

"We assumed that every household had seven individuals..." (huh? If asking how many dead people the house had, they didn't think to confirm how many live ones where there?).

Furthermore;
1) "We obtained January, 2003, population estimates foreach of Iraq’s 18 Governorates from the Ministry of Health. No attempt was made to adjust these numbers for recent displacement or immigration. "

This means that absolutely no attempt has been made to factor in the mass movement of refugees, including that of the large numbers of people back into the country since the fall of Baghdad to the coalition.

2). The article states in several places that the main violent killer in Iraq seems to be air strikes. If that is so, Human Rights Watch is telling a very different story, as I noted in an earlier post. And yet the article actually quotes HRW (on a different matter)in one of its opening paragraphs!

3). Most people in the military consider air strike activity to have peaked during the main three week offensive to Baghdad, after a pre-war build up of several months. In fact, contrary to popular belief, some of the larget air strikes- using up to 100 aircraft- preceded the ground war or even the official 'war' as such. In May last year the aircraft carriers where withdrawn and air activity dropped further. If the Lancet thinks the main cause of violent deaths is air strikes, how have they stayed at such a high level as the air activty diminished? And why where there not more civilians killed in the huge air raids from before the ground war? The article vaguely describes air strikes as 'helicopter gunships, rockets and other aerial weaponry' without attempting any sort of bomb dmaage assessment.

4). Contrary to what has been stated, the researchers did 'choose' samples. It says in the article that;

"During September, 2004, many roads were not under the control of the Government of Iraq or coalition forces.Local police checkpoints were perceived by team members as target identification screens for rebel groups. To lessen risks to investigators, we sought to minimise travel distances and the number of Governorates to visit, while still sampling from all regions of the country. We did this by clumping pairs of Governorates. Pairs were adjacent Governorates that the Iraqi study team members believed to have had similar levels of violence and economic status during thepreceding 3 years."

Like it or not, bias has been allowed to creep into the sampling process. In fact I think this part is trying to say that the interviewers themselves chose the samples, the wording isn't clear.

5). "When violent deaths were attributed to a faction in the conflict or to criminal forces, no further investigation into the death was made to respect the privacy of the family and for the safety of the interviewers."

In other words, the whole claim that deaths where largely caused by air strikes appears to be anecdotal. This also means that everyday murder is being counted as though it was always the coalitions fault, when it could be criminal. Oddly, enough, the article later conradicts itself by offering a breakdown of causes of death! Very odd...

6). "We estimated the death toll associated with the conflictby subtracting preinvasion mortality from post-invasion mortality,"

I can't find within where they got the pre-invasion mortality figures from (i'm curious here, have I missed it?). If the figures came from the old Iraqi ones then they would certainly undestimate violent deaths under the old regime, which would hardly have admitted to how it was behaving.

7). The article admites to hiring english speaking iraqis. Bearing in mind the problems with learning english under the old regime (it took government approval after the british council was removed in 1990, and the council only ever had two training staff anyway...) I'm not confident the researchers are un-biased. In fact the article itself says:

"Most importantly, the quality of data about births, deaths, and household composition is dependent on the accuracy of the interviews..."

8). Someone posted that mortality rates may have risen due to a sort of knock on effect from the damage casued by the war. And yet the article states that mortality due to non-violent causes is pretty much the same as pre-war. This doesn't seem to fit with the image of a nation suffering about 160-180 violent deaths a day by their figures.

When a combatant force takes up a position among a protected class (civilians) and continues to fight, the legal responsibility for the damage to civilians is theirs, not the attacker's.
BTW. Although our stuff is more powerful than it used to be, we use far less of it.
In Viet Nam, it would not be uncommon for artillery to fire ten thousand rounds between all the batteries involved, to support a rifle battalion, in a day or two.
One short news item in the last couple of days referred to a Marine artillery unit which had fired 150 precision rounds in four days.
We have an aircraft dropping a single bomb, instead of many aircraft dropping many bombs each.
We have helicopters using cannon or guided missiles instead of a salvo of unguided air to ground rockets.
This is a BIG difference even over GW1.

0. The assumption of 7 people per houshold was for the purpose of deciding how big a sample was needed.

1. They recorded whether there were empty households as a result of the residents fleeing or all being killed and there weren't very many so this would have little effect.

2. Most of the air strike deaths were reported in Falluja, so it is possible things are different there from the rest of the country.

3. The violent deaths in Falluja occured from April 2004 onwards.

4. That is not "choosing" samples. After pairing they made a random choice from the pair with the appropriate probabilities.

5. That is not a contradiction -- they are clear that they are reporting the word of the interviewee for the faction responsible.

6. They got pre-invasion figure by asking about deaths before the invasion.

7. Riight. Any Iraqi who speaks English must be a Baathist supporter. If that's the case we're screwed anyway.

8. The article does not say that non-violent mortality is the same. In fact, it says that about half of the 100,000 excess deaths were due to an increase in disease and accidents.

In full disclosure I supported this war to begin with but now believe that it was wrong-headed and is becoming ill-willed.

I also lay no blame at all on the US's professional military, whose senior officers opposed the war and whose serving men are bravely carrying out their orders to the best of their ability. I do blame the US's political executive for this mess-turned-monstrosity.

I have a few questions for The Real JeffS at November 16, 2004 at 12:21 AM and wilbur at November 16, 2004 at 03:06 AM:

1. How many Iraqis do you think have been killed in this preventive war?;


2. How many Iraqis do you think will be killed before a peace settlement is contrived?;


3. How many would be too many?

"The assumption of 7 people per houshold was for the purpose of deciding how big a sample was needed."

I'm beginning to understand now why nobody trusts statistics.

"1. They recorded whether there were empty households as a result of the residents fleeing or all being killed and there weren't very many so this would have little effect."

You've completely missed the point. The number of people moving back into the country is more than large enough to alter the nations pre-war and post war population figures, which in turn is yet another variable not considered that affects their calculations. They've assumed an identical population figure before and after.

The article ITSELF admits to this in one of many points where the authors state their are problems with the survey's accuracy;

"First,the use of government population estimates and the selection of households might have under-represented groups such as the homeless, transients, and military personnel."

"2. Most of the air strike deaths were reported in Falluja, so it is possible things are different there from the rest of the country."

The article considers the example of 61 violent deaths across the households involved, and this does not appear to be specifically Falluja or anywhere in particular. They state that three where due to coalition soldiers and the remaining 58 where killed by air strikes. The article clearly considers air strikes to be the main violent threat to civilains. This is in direct contrast to the investigation teams of Human Rights Watch who actually visited the scenes of many violent deaths and conducted investigations.

"3. The violent deaths in Falluja occured from April 2004 onwards."

No problem with this one. However, you haven't addressed the overall pattern of air strikes, which was at its highest 18 months ago, without a corresponding affect on the figures. And yet the article considers air strikes to be the main cause of violent death.

"4. That is not "choosing" samples. After pairing they made a random choice from the pair with the appropriate probabilities."

However, the original choice was made, not randomly selected, before this was done. When one looks at all of the variables and attempts by the article to state its own limitations, this is yet another thing dragging the accuracy of it down. To quote the article;

"Pairs were adjacent Governorates that the Iraqi study team members believed to have had similar levels of violence and economic status during thepreceding 3 years."

The decision on which areas had 'similar levels of violence' means human intervention, not random sampling.

5 Good spot- I've miss read the article. They have indeed worded that one as you say.

"6. They got pre-invasion figure by asking about deaths before the invasion."

Yes, but they did their calculation based on government figures. These in turn assumed a common population number before and after the war began. I'm also pretty certain that figures on pre-war population from the old regime will hide the huge number of violent deaths- the mass graves and so on- and even over state the size of population in order to avoid owning up to what the regime was doing. These stats where often fiddled by Saddam who even tried to claim half a million or more children had died from sanctions, where people have now come forward and admitted they where told to lie about the figures.

The whole figure of population comes into question. Coupled with the huge movement of refugees back into the country, it creates so many variables in all already inaccurate survey. And yet the survey has used this population figure to multiply the rate by in their calculations!

"7. Riight. Any Iraqi who speaks English must be a Baathist supporter. If that's the case we're screwed anyway."

Consider the affect on the data if only one of the survey group where dishonest. Now consider the affect if two where, and so on. They've selected half a dozen english speakers, can you be convinced that every single one of them is trustworthy?

To put it perspective, imagine going to Germany in 1946 and conducting a survey on mortality. Out of six english speaking Germans, could you be sure every single one of them had never had any Nazi sympathies ever? Now consider the affect if the Nazi's had banned the teaching of English without strict government approval for many years (I don't know if the Nazi's did, by the way, I'm theorizing).

"8. The article does not say that non-violent mortality is the same. In fact, it says that about half of the 100,000 excess deaths were due to an increase in disease and accidents."

I'll quote the article at this stage, which says:

"It is surprising that beyond the elevation in infant mortality and the rate of violent death, mortality in Iraq seems otherwise to be similar to the period preceding the invasion. This similarity could be a reflection of the skill and function of the Iraqi health system or the capacity of the population to adapt to conditions of insecurity."

There is yet more evidence that disputes the Lancet's findings.

Here's a UNICEF report on infant mortality in Iraq before the war;
http://www.unicef.org/newsline/99pr29.htm

If you read it, and then compare the figures with those provided by the Lancet, you'll find that;

1). The war has caused, according to Lancet figures, infant mortality to drop drastically today from those recorded by UNICEF in its pre-war study. In other words, bombing is good for them; Lancet's present day Iraqi infant mortality is HALF of UNICEF's pre-war figure.
2). The Lancet figures on pre-war infant mortality are completely out of whack with those offered by UNICEF. The Lancet estimates pre-war infant mortality to be about a quarter of what UNICEF actually worked out. UNICEF interviewed 40,000 households, the Lancet under one thousand.

Jack;
I don't profess to have a clue what casualty figures actually are; neither does the Lancet!
I don't know how many casualties are likely in the future either.

As for how many are too many- I think the losses of the US Civil War where uncomfortable to anyone in 1865. Measured against what the last 140 years would have been like, though, if the union lost, makes them seem less brutal. So it is with Iraq and the middle east. Blaming all the casualties on the coalition is simplistic, too. Many of these casualties are part of a long war by the Baathists on the Iraqi population that goes back decades. The only difference now is the proximity of western troops.

One other point that comes to mind; Tim stated "The study was peer reviewed before publication."

Yes it was- in about 7 weeks. That makes it peer reviewed much, much faster than normal.