Friday, October 13, 2006

Faith based surprise

So far, the only criticisms coming from the White House regarding the Johns Hopkins estimate of Iraqi war dead can be summed up as "That's a really big number, therefore it can't be true!" General Casey's criticism is only slightly more well-informed: "I haven't heard that number yet, so I doubt it is true."

Don't get me wrong, there's nothing wrong with skepticism, especially when a study comes along that challenges so many pre-existing beliefs. When confronted with the 650,000 deaths estimated by Johns Hopkins, our natural defense mechanisms tell us it can only be an outlier, a statiscal glitch on Satan's spreadsheet.

The best skepticism is informed skepticism. President Bush does not have the knowledge of statistics or sampling necessary to launch an informed criticism of the Johns Hopkins study. So he appeals to emotion, or the authority of those as scientifically ignorant as he. At yesterday's Rose Garden press Q&A, Bush harumphed something about faulty methodology, then followed with the customary regret that his elective war resulted in the deaths any innocent persons. From one so at war with science and fact-based decision making, Bush's explanation rings particularly hollow. Here's Billmon's take:
Well of course Bible Boy doesn't think it's credible. After all, what do Johns Hopkins University and The Lancet know about faith-based epidemiology? Nothing. They're just a bunch of doctors. Now if the study had been conducted by a committee of evangelical chiropractors from Oral Roberts University, that would be different.
The Johns Hopkins researchers used cluster sampling to arrive at a range of numbers, of which 650,000 is the median figure. Cluster sampling has been used for decades to sample large populations. If the methodology is wrong, then that would come as a shock to just about every insurance company on the planet, not to mention the Red Cross, Amnesty International, and the CIA. As usual, Majikthise does a great job of explaining a technical subject.

But don't wait for President Bush to hold forth anytime soon on cluster sampling, chi squares and regression to the mean. The Johns Hopkins study is just one more annoying little speed bump in the administration's road trip to glory. The only thing that will slow it down is a Democratic majority in the House and/or Senate.

1 comment:

Joseph said...

Those were my thoughts exactly. I read the abstract of the study and it looked sound on the surface -- speaking as someone who has by now studied a lot of epidemiology related to autism. Clearly, you can't assume that the only dead are those who have been reported in the media. The survey method is about as good as there is in trying to determine if there's an increased death rate after a particular date. To properly criticize the study, they need to either document methodological problems with it, or attempt a replication.