The Political Dismissal of Data
The Lancet, a prestigious British medical journal, released a peer reviewed article this week claiming that more than 650,000 Iraqis have died "as a result of the U.S. invasion and its ensuing chaos" (Euguene Robinson, San Francisco Chronicle, October 16, 2006). The methodology used was cluster sampling, which is often used for public-health research.
As expected, President Bush has called the study "not credible." Actually, what he said was "The methodology is pretty well discredited." We might also want to remember that Bush gave an off the cuff prediction estimate of "30,000, more or less" back in December of 2005, as reported by the Washington Post, and he was referring to Iraqi civilians, however in subsequent data, it was reported at 50,000. What is interesting is that the drama has been over that higher number, with many respondents indicating that the 50,000 is indicative of the time since December 2005, when the 30,000 number was indicated. So even if we acknowledge the 30,000 having reached 50,000, there's still a HUGE problem here, and that's the subject of this entry.
The reason given for the criticism of the clustering method used by Lancet is because it is too generizable, meaning that the sample error indicates the actual death level can be anywhere from 400,000 to 900,000. So, by focusing on this "error range" proposers of the 30,000 to 50,000 numbers can pretend that a number from thin air (the 30,000 one) is more "credible" than a number range based on clustering (low end of 400,000 to their high end of 50,000). Remember, Bush was pulling that number out of nowhere because attempts to get sources from him or his staff were unanswered. He pulled it out of the air, thinking that's probably how many that died. We use clustering samples every day in this country to predict and determine deaths in developing countries and we're so confident in the methodology that we allocate billions of U.S. dollars of funding based on it, yet when it says what we don't want it to say, it's considered "not credible."
Expect a long, bloody occupation that will have no "credible" end.
Stumble It!
As expected, President Bush has called the study "not credible." Actually, what he said was "The methodology is pretty well discredited." We might also want to remember that Bush gave an off the cuff prediction estimate of "30,000, more or less" back in December of 2005, as reported by the Washington Post, and he was referring to Iraqi civilians, however in subsequent data, it was reported at 50,000. What is interesting is that the drama has been over that higher number, with many respondents indicating that the 50,000 is indicative of the time since December 2005, when the 30,000 number was indicated. So even if we acknowledge the 30,000 having reached 50,000, there's still a HUGE problem here, and that's the subject of this entry.
The reason given for the criticism of the clustering method used by Lancet is because it is too generizable, meaning that the sample error indicates the actual death level can be anywhere from 400,000 to 900,000. So, by focusing on this "error range" proposers of the 30,000 to 50,000 numbers can pretend that a number from thin air (the 30,000 one) is more "credible" than a number range based on clustering (low end of 400,000 to their high end of 50,000). Remember, Bush was pulling that number out of nowhere because attempts to get sources from him or his staff were unanswered. He pulled it out of the air, thinking that's probably how many that died. We use clustering samples every day in this country to predict and determine deaths in developing countries and we're so confident in the methodology that we allocate billions of U.S. dollars of funding based on it, yet when it says what we don't want it to say, it's considered "not credible."
Expect a long, bloody occupation that will have no "credible" end.
Stumble It!
posted by Duane @ 12:20 PM
1 Comments:
At 2:48 AM,
Aufbau Ost said…
Good post!! Enjoyed reading it!
Post a Comment
<< Home