Statistics lie. But to claim they're useless is a neat way of evading what people do with them: stats don't 'become' useless, they're made useless. Real people distort the numbers, or ignore relevant figures. Which is why, when researchers use stastistics to make big claims - like the Lancet, claiming 655,000 dead in Iraq - they're so careful.



I won't defend the study directly: I think Lenin's Tomb demolishes the criticism quite effectively. Check out his defence of it here, and how he picks apart Iraq Body Count's (IBC) criticism here. And I'll include a discussion of methodology from his comment box later on. There is, however, another point to make.

Denying Holocausts

This is from the Institution for Historical Review (IHR), a Nazi front group, in their essay "What is Holocaust Denial?":

"Should someone be considered a "Holocaust denier" because he does not believe... that six million Jews were killed during World War II? ... Yet if that is so, then several of the most prominent Holocaust historians could be regarded as "deniers." ... Gerald Reitlinger, author of The Final Solution... estimated the figure of Jewish wartime dead might be as high as 4.6 million, but admitted that this was conjectural due to a lack of reliable information.

Never happened? Never Again.

Now, George W. Bush on the 655,000 Iraqi dead:

"The methodology is pretty well discredited," [Bush] said yesterday. Similarly, Iraqi government spokesman Ali al-Dabbagh told Reuters: "These numbers are exaggerated and not precise."

Mr. Bush has previously put the number of Iraqi deaths at 30,000. He reaffirmed that number yesterday. "I stand by the figure," he said. "Six hundred thousand or whatever they guessed at . . . it's not credible."

See the parallel? They dispute the overall figures, which lets them deny the event ever happened.


If the Occupation wasn't murdering Iraqis, there wouldn't be a resistance

Disputing the methodology is a pathetic response: would the Holocaust be less tragic if 4 million, instead of 6 million died? (Besides, historians put the figure closer to 15 million, once Communists, Slavs, homosexuals, the disabled and Jehovah's Witnesses are included.) But there's method to the madness. From the IHR's "Holocaust Remembrance - What's Behind the Campaign?":

The Holocaust remembrance campaign deserves scorn, not support, because it is a one-sided effort that serves narrow Jewish and Israeli interests and bolsters Jewish-Zionist power.

Holocaust revisionism is just a cover for anti-Semitism. And when Bush diverts the question from the form of the Occupation to its content, he's not saying he prefers 30,000 to 655,000. He's saying it doesn't matter how many have died.


Dead Nazi #19, by Marc Dennis

Once you deny the figures, it's a short jump to denying the event altogether. Which is exactly what the occupiers of Iraq want: for us to ignore the bloodbath they created. It's not the left's job to defend the Lancet's methodology. It's up to us to make the occupiers answer the charge: why have 655,000 died? Because there is no answer to that question, other than the amoral, brutal logic of capitalism.

Truths, Damned Truths and Statistics

Having said that, here's a defence of the methodology - the trainspotter in me finds statistics fascinating. From the report (.pdf) itself:

"The method, a survey of more than 1,800 households randomly
selected in clusters that represent Iraq’s population, is a standard tool of
epidemiology and is used by the U.S. Government and many other agencies."

From David Traiyner in Lenin Tomb's comment box, on methodology:

you're arguing against about 70 years of accumulated evidence that shows that random sampling, when appropriately safeguarded, works. It is an accepted and overwhelmingly validated technique... It is therefore hard for me to assume that you're not simply grasping at straws because you don't want to believe that 650,000 is plausible....

I suggest you read some of David de Vaus's work on the matter. 'Surveys in Social Research' is a pretty good introduction and, although it focuses on political science, also lays some of the groundwork for general statistical methodology. Or, if you prefer, Sampling Methodologies with Applications by Rao Poduri.

On sample size:

it is not a 'pathetically tiny sample' it is a very good sample size, well with the safety margin and approved by four independent authorities during the peer review process. 1,849 households containing 12,801 persons allows 95% probability of obtaining the correct result. That is not something plucked from the air but a standard technique that has been consistently validated over decades. It is the standard technique of countless governments, NGOs and universities and, put simply, it has been proven to be very accurate.

The whole point of a representative same is that it *represents* the whole -that's why we call it representative. It doesn't mean representative merely in the sense that one thing can denote another: it means that the properties of the sample can be reliably assumed to be the properties of the whole.