Anybody Know Any Christian Statistician Jokes?

We start with a bang… with Robert Fisk’s allegations of Israel having used a uranium bomb in this year’s war with Lebanon. I wonder what the moral distinction is between a battlefield “uranium bomb” and the terrorist “dirty bomb” so feared by the talking heads?

We continue with this article courtesy of EK Hornbeck. It’s actually the transcript of a speech given by torture victim Maher Arar. Definitely worth reading to get a full understanding of his ordeal and of the guilt of both the American and Canadian governments. I will never forget meeting a foppish middle-aged housewife here in Ottawa, the spouse of an acquaintance of a famous (yet unnamed) Canadian author who is a friend of mine; the housewife exclaimed, quite unprovoked, “Well that Arar fellow must have been up to something. Why else would they have tortured him?” That, my droogs, is the level of ig’nance we face.

Good news: the Public Health Strengthening Project in Guyana, the one on which I consult, has been shortlisted for a major award called, “the Canadian Award for International Cooperation”! We find out on Tuesday if we’ve won!

So let’s talk a bit more about that Lancet article which estimates 650,000 have been killed in Iraq as a result of the war. On a comment on another blog, someone had written that the number seemed unrealistic since the UK had endured the same order of casualities after several years of Nazi bombing during WWII. What people need to realize is that the Lancet study purports to measure all downstream casualities of war, not just those killed by bombs and bullets. That means it includes deaths by malnutrition, disease, random violence, suicide, etc, if they can be linked to the invasion/occupation. This strikes me as a very proper way to conceptualize the violence of war, especially in the modern context wherein infrastructure destruction is so easy, so prevalent and has such a devastating long term impact on quality of life.

Here is a more detailed discussion of the study’s controversy. Meanwhile, the media has made a big deal about a Statistics Canada statistician named Scott Gilbreath having “debunked” the Lancet study on his blog. Gilbreath is a self-described “Perpetually perplexed Christian statistician” and his political posts are unsurprisingly pro-neocon. That’s his business, and he’s entitled to it and to his opinions; I begrudge him nothing. Good for him for not shirking from his identity.

Interestingly, though, Gilbreath’s most popular post, titled “Lancet study of Iraqi deaths is statistically unsound and unreliable”, is missing at the time of this blog post (2:AM Monday morning). This missing post is purportedly chock full of juicy analyses tearing down the Lancet study’s methodology. So why has it been pulled? Anyone? Bueller?

You will recall from my last post that “main street bias” has been the charge lain by the study’s detractors and by the conservative echo chamber. Well, there’s another bias at play in the study, a convenience sampling bias resulting from security concerns preventing researchers from entering dangerous areas. Interestingly, the result of this bias is to underestimate the death toll due to war-related violence.

So if these detractors are so keen on exposing the poor methodology of s scientific study, why have they only chosen to discuss a bias toward overestimation? It wouldn’t be because –gasp!– they have a pro-war agenda to prop up?

The simple fact is that every population study is rife with methodological errors and compromises, some of which will bias results in one direction, while others will bias results in the other direction. The question no one has yet asked is, which bias is more potent and thus has more influence over the reported results?

I remain favourably disposed toward the study. I’m impressed by its creativity and rigor, given the difficulty of the study environment. The issue of insufficient test clusters is discussed in more detail by the Christian statistician here, and I’m unsure of what to think about it. It seems to me that if his concerns are largely a sample size issue, then the wide confidence interval accounts for some of it.

You hardcore statisticians out there…. I know you’re reading. Pipe up, already! This is your chance to contribute to some important shiznit!