OxBowBusiness

Statistics and not Politics

You might have seen the latest spat over how many people have, tragically, been killed in the current Iraq conflict. Estimates seem to vary from around 100,000 to as many as 943,000. What's the truth and how do they know?

A very short article in the latest edition of the Economist provides at least part of the answer as to where the differences in estimates come from. I can't quote the entire article of course and the salient points are these:

"COUNTING exactly how many people have died as a consequence of war is something of a dark art. This is particularly so in Iraq, where various approaches—including hospital death data, mortuary tallies and media reports—produce different results. A study based on statistical techniques by researchers at Johns Hopkins Bloomberg School of Public Health, in Baltimore, has concluded that many more people have died than revealed by these other methods. It claims that an extra 2.5% of the Iraqi population has died since the country was invaded in March 2003, mostly as a result of violence.

The study, published in the Lancet, is based on random sampling. Selecting small numbers of people at random allows statisticians to say something about the whole population ...

So the researchers used a technique that is called clustering. Clustering works by picking out neighbourhoods at random and then surveying all the people living in them. Gilbert Burnham and his colleagues gathered information on deaths from 47 neighbourhoods, each containing almost 40 households."

If youa re interested in statistics and so on, then this article is well worth a look: page 73 of the print version and a subscription is needed to access the article on line.

Duncan Williamson

posted by Duncan Williamson @ 6:37 AM