Say Anything Blog

There are serious questions about the methodological and ethical practices of of Roberts and the other researchers involved in the Lancet study.  I put zero weight on that study because of their sloppiness has made the quality of their data untestable.

However, there are some questions about this study too.

They multiply their final numbers by an arbitrary 35%.  What rubbish.

They don’t separate out civilian deaths from military/insurgent deaths.

I’ve yet to see the demographics of the casualties and whether this follows the general demographics of the population (a way to tease out military from civilian deaths).

Nonetheless, even with what appears to be a high ball number compared with other measures of the Iraqi death toll, it does put to bed the question of whether the Lancet study should be taken seriously.

It clearly should not.

That 35% was for

underreporting of deaths due to migration, memory lapse, and dishonesty.

We could for example go with Iraq Body Count’s lower figures of 81 to 88 thousand violent deaths.

Do not fear to be eccentric in opinion, for every opinion now accepted was once eccentric.

MOFAL, I know their justification for it.  It’s like multiplying your statistical error bars by the square-root of pi, like they used to do in the early 20th century, before they understood how to write down error budgets that included systematic uncertainty.

Migration (emigration as well as immigration), memory distortion and dishonesty could cause the numbers to flex either way.  The story is always better with each telling.

While I respect the IPC numbers, it has its problems too.  (What metric doesn’t in a war zone?) For example, there is also concern that they have mixed in insurgent/militia deaths with the generally populace. 

However, they haven’t provided demographic information on their full data set either, which would allow one to ascertain the degree to which this is a problem.  That like problems with other surveys is a problem that is readily fixable.... by simply providing the data. 

Really what’s the problem with these guys?  Why can’t they follow established research guidelines?  Why is that so difficult?