Statistics vs. Politics: Estimating Deaths in the Iraq War.

As many folks know, my background is in quantitative statistics. When the Lancet came out with a second study on the number of deaths in Iraq, George Bush dismissed the report stating, "I don't consider it a credible report...the methodology is pretty well discredited." George Bush's response fascinating me since the Lancet is one of the most well-respected medical journals in the world and it seemed quite unlikely that it would publish scientific analysis that didn't pass peer review, and the peer review is pretty damn rigorous.

So I decided to take a look myself at the Lancet report. First, the authors used a random sampling of 50 "clusters" (controlling for population density) that would provide both the statistic power and 95% confidence interval to determine whether there was, at least, a doubling of the death rate post-invasion. In each of the 50 clusters, 40 households were interviewed by medical doctors.

The results from the interviews were quite stunning, and the information was broken down in Table 2 of the study. Most stunningly, while violent deaths composed 2% of the total deaths pre-invasion, violent deaths made up a full 55% of total deaths post-invasion.

Read more...

The log-linear regression models utilized by the authors helped control for the correlation of deaths within the same cluster. While this analysis "assumed that the variation in mortality rates across clusters is proportional to the average mortality rate," additional analyses were conducted to allow base death rates to vary by cluster.

Three of the 50 clusters were dropped in the final analyses due to logistical problems in sampling populations there, so the sample size was 1849 households in 47 randomly selected clusters. Surveyors asked for death certificates in 545 (87%) of the 601 reported deaths; and these certificates were provided by respondents in 501 of these cases (92% of the time).

The report's findings were that "the crude mortality rate in the pre-invasion period was 5.5 per 1000 people per year (95% confidence interval 4.3–7.1) and for the overall post-invasion period was 13.3 per 1000 people (10.9–16.1) per year." These discrepancies had p-values of <0.0001 -- which provides ample evidence that this increase is not likely to be random variance. Here's the geographic breakdown:

Whereas violent deaths skyrocketed post-invasion, no significant different in non-violent deaths (p = 0.29) was found. But based on pre- and post-invasion mortality rates, the authors conclude:

We estimate that between March 18, 2003, and June, 2006, an additional 654,965 (392,979–942,636) Iraqis have died above what would have been expected on the basis of the pre-invasion crude mortality rate as a consequence of the coalition invasion. Of these deaths, we estimate that 601,027 (426,369–793,663) were due to violence.

With all this in mind, I'd like to know who is discrediting this report and on what basis -- because the results look both quite damning and rigorous. As someone with a far better statistical background than George Bush, I'm curious upon what basis he is dismissing this scientific research.