<HTML>
<HEAD>
<TITLE>Re: [Media-watch] exhange with Blairs private secretary</TITLE>
</HEAD>
<BODY>
<FONT FACE="Verdana">Hi David<BR>
<BR>
You have probably already seen this but in case not here is some of what the Lancet report authors have said in emails to medialens – there is more at http://www.medialens.org/. <BR>
<BR>
Regards<BR>
<BR>
John<BR>
<BR>
<FONT SIZE="4">We asked the report's authors about the large rise in numbers of estimated civilian deaths over previous estimates, and also on the ability to make a reliable body count without bodies. Dr. Gilbert Burnham responded:<BR>
<I>"In short, we used a standard survey method that is used all over the world to estimate mortality. So bodies are not necessary to calculate mortality. In fact going to the community for household surveys on mortality is the standard method used for calculating mortality all over the world, and is probably the method used in the UK census as well, although I am not a demographer. <BR>
"Anyway, information collected in surveys always produces higher numbers than 'passive reporting' as many things never get reported. This is the easy explanation for the differences between iraqbodycount.net, and our survey. <BR>
"Further a survey can find other causes of death related to public health problems such as women dying in childbirth, children dead of infectious diseases, and elderly unable to reach a source of insulin, which body counts cannot do--since they collect information from newspaper accounts of deaths (usually violent ones). Can one estimate national figures on the basis of a sample? <BR>
"The answer is certainly yes (the basis of all census methods), provided that the sample is national, households are randomly selected, and great precautions are taken to eliminate biases. These are all what we did. Now the precision of the results is mostly dependent on sample size. The bigger the sample, the more precise the result. We calculated this carefully, and we had the statistical power to say what we did. Doing a larger sample size could make the figure more precise (smaller confidence intervals) but would have entailed risks to the surveyors which we did not want to take, as they were high enough already.<BR>
"Our data have been back and forth between many reviewers at the Lancet and here in the school (chair of Biostatistics Dept), so we have the scientific strength to say what we have said with great certainty. I doubt any Lancet paper has gotten as much close inspection in recent years as this one has!"</I> (Dr. Gilbert Burnham, email to David Edwards, October 30, 2004)<BR>
<BR>
Channel 4's Tom Clarke had made a further observation:<BR>
"The definition of civilian is also unclear. The majority of violent deaths were among young men who may - or may not - have been insurgents."<BR>
The report's lead author, Dr. Les Roberts, responded to this point:<BR>
<I>"The civilian question is fair.? About 25% of the population were adult males.? >70% of people who died in automobile accidents were adult males.? Presumably, they died more than other demographic groups because they are out and about more.? 46% of people reportedly killed by coalition forces were adult males.? Thus, some of them may have been combatants, some probably were not... perhaps they were just out and about more and more likely to be in targeted areas.? We reported that over half of those killed by coalition forces were women and children to point out that if there was targeting, it was not very focused.? Thus, we are careful to say that about 100,000 people, perhaps far more were killed.? We suspect that the vast majority were civilians, but we do not say each and every one of the approximately 100,000 was a civilian."</I> (Email to David Edwards, October 31, 2004)<BR>
<BR>
</FONT><BR>
</FONT><BLOCKQUOTE><FONT FACE="Verdana"><FONT SIZE="1"><BR>
<BR>
</FONT></FONT></BLOCKQUOTE><FONT FACE="Verdana"><FONT SIZE="1"><BR>
</FONT></FONT>
</BODY>
</HTML>