Thursday, February 14, 2008

A 'blogswarm' around a made-up number

Apparently the anti-war bloggers are having a "blogswarm" to commemorate the 5th Anniversary of the Iraq Invasion. I do love there completely made up button though (see image on right). It's that little old chestnut of an extrapolate made-up figure of "1 million" dead which is based upon a report in Lancet that did a survey of a 1000 people, then expanded it out through linear extrapolation.

Of course, the anti-war lobby loved it, and, as they are such immensely bright and superly clever people they decided to carry on the calculation on the assumption that it represented reality and now proudly display an inherently flawed and unfalsiable figure as a fact - sometimes with an increasing ticker too for added effect!

Immensely bright and clever people who are actually complete idiots that scream about propaganda and then display equally bullshit propaganda as some sort of scientific reality. They're just like the climate change campaigners that say that a "peer-reviewed" document is an unequivocal fact. Funnily enough they tend to be clever and bright lefties too.

32 comments:

Old BE said...

There is a band called Million Dead but I think they pre-date the Iraq invasion.

Bob Piper said...

Funnily enough they tend to be clever and bright lefties too... like that old leftie campaigner John MacCain.

Pete Chown said...

At the moment, the Lancet figure is the best estimate we have, so we have to use it. If someone comes up with a better estimate, we'll drop the Lancet figure and use the new one instead.

Very few things are backed by perfect evidence, so if we're being pedantic, we can't actually be certain about very much. We can only go with the theory that is backed by the best evidence, while being honest about its limitations.

dizzy said...

That's simply not true about it being the best estimate. They asked 1000 people if they knew anyone who died a violent death since the invasion and then that was equated to being caused by the invasion. Its bollocks.

Pete Chown said...

I just realised that the one million figure comes from the ORB poll, not the Lancet paper. The Lancet paper suggested around 650,000 deaths resulting from the war.

The Lancet paper looks at the increase in violent deaths since the invasion. They're not attributing all violent deaths in Iraq to the war. They do, however, count deaths which are not caused by military operations. For example, if deaths due to armed crime increase, they would count that increase as resulting from the war. This isn't right or wrong, but you have to be careful that the study is answering the specific question you want to ask.

What estimate would you prefer, if you don't like this one?

dizzy said...

No estimate.

Anonymous said...

Dizzy - you are wrong on this one.
The Lancet figure was quietly accepted by the government 6 months after dishing it.

If medical statisticians and the govt can accept it then why not you?

dizzy said...

I'm a statistician and the method is dodgy as hell. Just because some peiople accept something it doesn't make it right. Lots of people used to acceopt that the earth was flat, how right they were. So no, I;m not wrong, I'm right for saying the figure is made up bollocks. You can draw a statistical comparison between the reduction in piracy and global temperature as well.

Pete Chown said...

It's true: there is a correlation between the fall in piracy and the rise in global temperature, but one doesn't cause the other.

The Lancet paper found that violence was a factor in 2% of deaths before the invasion, and 55% afterwards. You're right that, strictly speaking, the study doesn't prove that the war caused this dramatic change. But we have to use our common sense about this, don't we? What else might have caused such a big rise in violence?

dizzy said...

a butterfly flapping it's wings in Mexico?

Anonymous said...

The original figure of 650,000 was extrapolated from only 600 deaths that could be proved to have been caused by the conflict.

That's like considering the 90 deaths from c.diff at Maidstone NHS Trust and deducing that the NHS is killing 97,500 people per year through c.diff alone.

IIRC, 650,000 is more than double the total number of military and civilian deaths for the UK and Commonwealth for the whole of the second world war.

As has been said before - it's bollocks.

Anonymous said...

For lefties, intellectual honesty, especially when it comes to statistics, has never been important. What matters is to feel comfortable as a member of the tribe. That's why, when Mao was murdering 70 million people, you had to wear a Mao badge. Nothing to do with China, just, 'Hi Guys! I am one of you.'

Anonymous said...

I must learn to spell my name.

Pete Chown said...

Using the Maidstone figures for the whole NHS would be like using the Falluja figures for the whole of Iraq. If you look at the paper, you will see that they didn't do this. They took samples from violent areas and comparatively peaceful ones.

That isn't proof that the samples were chosen in an unbiased way, but if you think they weren't, you need to be specific about the problem. Speculating that there might have been bias doesn't really help, because anyone who has done any statistics is already aware that sampling bias is a potential problem.

Anonymous said...

"Using the Maidstone figures for the whole NHS would be like using the Falluja figures for the whole of Iraq."

That's essentially exactly what was done. I read the report when it was published.

There are too few samples to allow a reliable, straight line extrapolation to take place for a country the size of Iraq.

In addition, the sample points were disproportionately taken in urban areas in the south of Iraq. The rural, Bedouin population and the Kurds in the north, who were largely untouched by the fighting, were conveniently ignored.

The author of the report is a non-practising medical doctor who has no training in statistics other than the basic, medical statistics courses that are part of an undergraduate medical degree. He has a history of publishing 'academic reports' that are very critical of the US government.

This is a classic example of an unqualified 'academic' deciding on the outcome at the start of his 'research' then trying to find evidence to support his assertions.

Do you really believe that the number of Iraqis killed is nearly quadruple the number of UK and Commonwealth citizens killed during the whole of WWII?

Andy said...

Ok Pete - how does this fit with your narrative: http://hurryupharry.bloghouse.net/archives/2008/01/10/new_study_estimates_lower_iraq_death_count.php

Anonymous said...

If you want a fairly lengthy article detailing the design of, problems with, and criticisms of, the Lancet II report there is this article. There are plenty of links within the article to follow up for more details.

There is another another survey, this time from the Iraqi government and the World Health Organisation, that puts the figure at about 151,000.

Anonymous said...

Thats right Dizzy, just stick your fingers in your ears and say over and over "The Iraq war was a brilliant idea, I havent been proved wrong" over and over.

Are you ever going to accept the Iraq war has been a complete fuckup and a massive mistake for the UK to take part?

(PS I am not a leftie)

dizzy said...

What's this got to do with whether it was a policy mistake or not? This is about a crappy figure being crappily used.

Anonymous said...

Yeah cos the Lancet is so left wing Dizzy

dizzy said...

oh gfor fuck sake, another anonymous straw man. Did I say the Lancet was left wing? No, I said the twats using the figure were.

Pete Chown said...

The Lancet study lists the provinces where sampling was carried out, and also the populations of those provinces. You can see that there isn't visible bias between violent and non-violent provinces. Anbar, probably the most violent, gets three clusters; Baghdad gets twelve. Both these are reasonable bearing in mind the populations.

As a separate issue, it's possible that too much sampling was done in towns. The ORB study was criticised for this, and their latest one includes more rural areas. They found that, in the event, little bias had been introduced by the original urban focus.

The Lancet and the ORB studies are in the same ballpark. The Lancet study suggested 400,000 to 900,000 and the ORB study suggested 700,000 to 1,400,000. The Iraq Body Count will be an underestimate because it only counts deaths which get reported in the media. So the odd man out is the WHO/Iraqi government survey that put the figure at 150,000 deaths.

Of course I don't know for definite why the WHO survey came out with an answer that was so different. One reason might be the fact that interviewers were officials from the Iraqi government; this could have influenced the answers people gave.

There was also a problem in that some areas were considered too violent for the government researchers to visit. Figures were imputed to these areas using ratios from the Iraq Body Count. So for example imagine the IBC showed that X-Town was twice as dangerous as Y-Town, and X-Town was too dangerous to visit. They would impute a figure to X-Town which was twice what the survey found in Y-Town.

The problem with this strategy is that X-Town is probably too dangerous for journalists to visit, as well. As a result, while the IBC figure will be an underestimate everywhere, it will be especially so in the more violent areas. Imagine X-Town as Falluja and the problem becomes particularly clear. Virtually no outsiders go there, so deaths are particularly likely to be unreported.

Chris Paul said...

I agree with Dizzy on this. The Lancet figure (singular) is very very suspect. Not least because the Lancet had a humungous RANGE of possible figures, which fact is never reported.

All the likely errors from the methodology are in one direction, and everyone involved admits that the sample is way way too small even for the normal confidence levels in such an exercise.

If we take Iraqi Body Count - erring on the low side - and the lower half of the Lancet range we get in the right ball park in my view.

Although I get in some trouble with some anti-warrers over my denial of the Lancet headline figure, and would too for any comments I chose to make about Gulf War Syndrome, it strikes me that in terms of the principle of this war and the wider war on terror the numbers just do not matter.

Why are anti-warrers wishing the figure up? Determined that Mr Rusling's figures for GWS (this is Iraq 1 btw) should be right?

Quarter of a million, half a million, one million. Makes no difference to the arguments for or against the war.

paul ilc said...

Game, set and match to Dizzy.

paul ilc said...

Why are anti-warrers wishing the figure up?...Quarter of a million, half a million, one million. Makes no difference to the arguments for or against the war.

Quite, CP. Propaganda is the answer, as you implicitly suggest. Leftists often deliberately conflate consequences and intentions, not to mention ends and means, to further their ends. As you imply, the numbers argument is a grossly utilitarian/consequentialist one. If the war was wrong in principle, then just one death would be a moral outrage. If the war was right in principle, any deaths would be tragic but the lesser of two evils.

Of course, morally, (foreseeable) consequences matter (up to a point), but so do principles and intentions.

Alex said...

The Lancet figure could well be suspect, but I am surprised that there isn't another back of the envelope calculation that would give an idea of whether it is in the right ballpark, or across town from the ball park.

I remember getting really annoyed with Blair et al. after 9/11 when they were to my mind playing up the number of British casualties in the WTC. They started out at 700, and it gradually fell day by day to around 200, then they stopped announcing the figures quite so loudly until the number dropped to something err. under 100 then they actually put 72 on the memorial at Grosvenor Square, but looking at the list it turns out that a good proportion of those had tenuous links to the UK - long term emigrants who had taken US passports, UK born children of US expats who had gone back to the US years ago, Ugandan Asians who carried a US passport but had no nexus to the UK, Bermudans etc, etc. What was particularly galling was that it was pretty obvious to anyone who had spent a lot of time in the US (I used to live there a long time ago and made several week long trips to NY in 2001) that it was utterly implausible that 15% of the 5,000 people killed in the WTC would be British (absent the NY office of a British bank being in the WTC, which they weren't). It is hard not to notice the occasional Brit on the streets of NYC or in America, but 15% - gimme a break.

For the PM and his cabinet to give such a high figure could have been a question of incompetence, but in retrospect is more likely to have been wilfull exaggeration.

Bob Piper said...

Pete, I suspect dizzy won't be back. Cheap rhetoric is so much easier than having to respond to detail.

Couldn't you make it more 'Daily Star' for him in future please, it makes a 'statisticians' life so much easier when he is pretending to be a politician.

dizzy said...

A politician? Why would I want to be a politician? That would mean I would have to betray the principles I have and become a hypocritical compromised arsehole Bob. Presented company included of course.

At the end of the day though BOb, you have to admit there is something deliciously ironic about a comment which suggest I am all about rhetoric rather than detail which then proceeds to plays pretty lame rhetoric whilst ignoring the detail.

The fact that Chris Paul even gets it, suggest that my post, rather than itself being rhetoric, is actually attacking the idiotic rhetoric of the anti-war movement in relation to the made-up figures they push. Something which I note you have avoided commenting on, instead going for rhetoric.

Mind you, I'm also surprised Justin hasn't turned up to say how important it is that a million have died. After all, it was his blog that led me to the nutter. Him and Ireland are equal idiots in banging on about the made up number. some might say Tim does it in an "obsessive" way. but then I would only be saying that for tactical and malicious advantage!

I love Base64!

Elby the Beserk said...

"It's that little old chestnut of an extrapolate made-up figure of "1 million" dead which is based upon a report in Lancet that did a survey of a 1000 people, then expanded it out through linear extrapolation."

Just soze you know, the govt trounced those figures, and then some weeks later, VERY quietly (after all, they are as we know, serial liars) agreed that the method used by Lancet was rigorous and correct. Not, as you suggest, bullshit propaganda.

Don't tell me - the invasion of Iraq has been a triumph for democracy yes? Jeez, you'll be telling us Brown is a hero next.

ERGO the projected number of dead was reasonable. Add that to the 1/2 millions lives lost because of sanctions before the war.

The collateral from the disaster in Iraq will have seen many dead who never reached any place they might be "counted". Or we take the guy mown down by a tank, who was listed as a "traffic accident". And sanctions were still in place of course, so no aid, no medical help was getting to the poor Iraqis.

WE have devastated this country. In the name of freedom.

Check you facts more carefully, Dizzy - you look idiotic. And regardless, the loss of lives thought our venal intervention in that country is shocking, and Blair should be tried.

http://preview.tinyurl.com/23zszv

Unknown said...

"I do love there completely made up button though (see image on right)"

Thanks for loving the graphic ('their' typo?) which indeed was made up although the figures are not imaginary. I am one of the 'twats' you excoriate (the idea belongs to a fine fellow from New York blogging as GodlessLiberalHomo), I designed that and most of the other badges incidentally. I am sufficiently persuaded by both Lancet studies and both ORB surveys as well as the detailed discourse the authors have engaged in to explain their work. I am therefore confident this is not 'bullshit propaganda' nor is it repeated without thought or research and the weighing of contrary arguments. One should also not forget the refugees, both internally displaced people and those who have managed to flee Iraq number in the millions (UN figures show 2.3 million IDP's & 2.5 million outside of Iraq) not only is the UK not admitting its share it is repatriating Iraqis to unknown fates.
Still, thanks for the link and good luck with that 'Thinks' effort.

libhom said...

The Lancet study was a peer-reviewed study in a prestigious scientific journal. It is important for people to accept facts rather than to deny them because they don't fit a certain political agenda.

I encourage members of the Reality Based Community to join the March 19 Blogswarm Against the Iraq War.

libhom said...

Doctor Syn:

The "Iraqi government" has been under strict orders to minimize death statistics since shortly after the invasion.