Wednesday, November 17, 2004

100,000 or 300,000

Back to those extra deaths in Iraq which were discussed earlier. After reading the Lancet report itself, I can understand some of the comments made at Harry's Place and look at some of the figures with a bit more precision.

'During the period before the invasion, ... the crude mortality rate was 5·0 per 1000 people per year'

After the invasion, from March 19, 2003, to mid-September, 2004, if the Falluja cluster is excluded, the 'mortality is 7·9 per 1000 people per year'. So, in a population of 24.4 million, the additional 2.9 per thousand equates to 71,000 people per year or for a period of 17.8 months a total of 105,000 deaths.

If the Falluja cluster is included, 'the crude mortality rate...  was 12·3 per 1000 people per year'. So, the additional 4.4 per thousand equates to 107,000 people per year or a total of 159,000 deaths.

This is higher than the 'doubling' - around 100,000 - that was floated on the radio, but lower than the report's 'point estimate of about 200 000 excess deaths in the 3% of Iraq represented by this cluster.' That 3% of Iraq would be about 732,000 people.

So, 159,000 deaths in a population of 732,000 ?

In any further analysis I risk making basic statistical errors such as drawing conclusions from to small a sample, but looking at when the deaths from the Falluja sample are supposed to have occurred, roughly from the histogram of Figure 2 : in 2003, Jun - 3, Nov - 1 ; in 2004, Apr - 11, Jun - 4, Aug - 31, Sep - 5 (adding up to 55, whereas the real total from the sample was 53, 52 of them violent).

31 deaths in the sample for August 2004 ? That would imply 93,000 deaths.

For April 2004, 11 deaths implies 33,000 from the Falluja sample or, scaling it down, 13,500 for the city of  Falluja itself (pop. 300,000).

Comparing it with the figure from the NYT of 850 casualties from the April fighting, it is multiplied about 15 fold, as against sixfold when the Lancet report is compared with recorded figures for the whole of Iraq.


Post a Comment

<< Home