Thursday, October 12, 2006

The Latest Lancet Survey of Iraqi Deaths

A lot of heated atmosphere (hot air) is being generated by the Lancet Study claiming 600,000 excess deaths in Iraq since Saddam stopped being numero uno.

In my opinion this survey depends on dubious sampling methods.

Why not just go to the Iraqi government for the real figures? Or go to a more authorative source, Iraq Body Count?

If we extrapolate backwards from the 3,000 (actually less) deaths reported last month and assume that number was constant (actually it was lower).

You get 36,000 deaths a year or about 120,000 in 3 years.

That is 1/5th the number this study reports. (note: Iraq Body Count puts the number at between roughly 44,000 to 49,000 for 3 years making the Lancet study off by a factor of at least 12).

Real engineers do reality checks. Rough order of magnitude stuff. Real engineers always look to see if prejudice affects the numbers. Or "are we in the right ball park"?

This study does not pass the smell test.

H/T Instapundit

Also see Tim Blair who's commenter rhhardin (18th comment) did a gedanken experiment about sampling cats with sharp needles. He found that after 1 to 4 samples further sampling showed that cats were mostly teeth. Children, this is a job for professionals. Do not try this at home.

And more Tim Blair


Update: 19 Oct '06 1908z

A critique of the sampling methods used in the survey by a statistician who has actually done surveys in Iraq. The big flaw? Small sample size. A margin of error of 1200% gives away the game.

No comments: