Tuesday, October 02, 2007
Analyzing civilian casualties in Iraq
If you read only one blog post today, see John Wixted's tour de force analysis of civilian casualties in Iraq. Using the data from the anti-war Iraq Body Count, he shows (i) that IBC data corroborates General Petraeus (which fact IBC declined to acknowledge), (ii) that the Moqtada al-Sadr's Mahdi army is by and large working with the United States, rather than against it, and (iii) that al-Qaeda's suicide bombings are almost certainly intended primarily to influence American public opinion rather than the political or military conditions in Iraq.
CWCID: Glenn Reynolds.
11 Comments:
By 
David M, at
Tue Oct 02, 11:31:00 AM:
Trackbacked by The Thunder Run - Web Reconnaissance for 10/02/2007
A short recon of what’s out there that might draw your attention, updated throughout the day...so check back often.
By 
, at
Tue Oct 02, 11:36:00 AM:
"Meanwhile, I've been digging into the IBC numbers, and what I've discovered is simply amazing. I'm going to show you two incredible charts today, both of which confirm what I've long thought to be true."
lost me right there, when he abandoned any pretense of objectivity. Why even bother with the analysis when you've stated your conclusions ahead of time?
By SR, at
Tue Oct 02, 02:43:00 PM:
phrizz,
If you follow Engram, the conclusions come from intense study of raw data over a prolonged time span. These are conclusions the kind one can make with a bit more confidence than that which one might have after reading, say the New York Times. If you read the blog regularly, I feel certain you could rely on the notion that data out of line from the series he has developed wouldn't be tossed off, but rather examined to see if there was a legitimate reason for the difference.
By Dawnfire82, at
Tue Oct 02, 06:31:00 PM:
"Why even bother with the analysis when you've stated your conclusions ahead of time?"
I wonder that every time I hear about some new report on global warming, multiculturalism, civil liberties, abortion, and so forth.
If, by chance, they might have happened across results that run counter to their ideas, they just won't report them.
By , at
Tue Oct 02, 09:36:00 PM:
"lost me right there"
yeah, why bother to critically analyze the argument - let alone read it - when you can just search for a reason to dismiss it out of hand?
By Purple Avenger, at
Tue Oct 02, 10:04:00 PM:
Why even bother with the analysis when you've stated your conclusions ahead of time?
Suspicions != "conclusions"
By , at
Tue Oct 02, 10:52:00 PM:
moqui: Some parts of his argument I agree with: was Petraeus' report of casualty numbers accurate or at least uncooked? Yes, it looks like it, so he seems credible in that respect. Other points are stated without proof: he asserts causes for the decline in execution-style killings and for the increase in suicide bombings during the surge, but doesn't present any proof for this statement, only showing that the data is consistent with his assumptions (which is dishonest but a common rhetorical device).
I certainly don't accept his point based on his projection for September's suicide bombings, because the data is derived from a different source using undeclared methods. I think it's far too early to decide whether or not the surge has made civilians safer, a) because there's no real september data yet, and b) there is no analysis is presented as to whether or not the trend in killings (which is certainly positive up to and including August during the surge) could be due to random monthly fluctiations in the number of deaths.
There, I did the thinking for you. Happy?
By SR, at
Wed Oct 03, 02:03:00 AM:
Phrizz,
You must have missed Engram's presentation of the data as a trailing moving average of the previous three months. The trend is the same when "random monthly fluctuations"are taken into account.
Why should we accept your conclusion that you are thinking?
By Fritz, at
Wed Oct 03, 10:43:00 AM:
This comment has been removed by the author.
By Fritz, at
Wed Oct 03, 12:03:00 PM:
SR: What? As far as I can tell, the author is trying to make a statement about whether or not the troop surge has coincided with a reduction in the number of civilian deaths. He presents no model for how deaths are distributed, so it's difficult to interpret the results.
Assuming an average pre-surge death rate of 1000/mo for execution style killings, after the surge, after the surge from the graph it looks like the purple bars add up to about 4,100 dead in 6 months. The expectation value based on the pre-surge rate would be 6,000 dead in 6 months.
Now you have to assume some distribution function for deaths and look at the cumulative distribution function if you want to know what the probability of this occurrence would be. Certainly, if the data are Gaussian, and assuming sigma = 400 deaths (rough guess from the graph), then there is only a very small chance that this is a random fluctuation. I'm sure that the # of deaths is not Gaussian distributed and that this is kind of a silly guess, but I don't have the time or energy to try and figure out what the right distribution is. The point is you need to do this kind of analysis if you want to know how likely this is to be a fluctuation.
So yes, SR, I am at least trying to think, and not just regurgitating something someone else said.
By , at
Wed Oct 03, 02:44:00 PM:
I thought the point of the surge wasn't to lower civilian deaths while the surge is going on, but rather to buy time for various political gains, so that when we go back to previous levels (or withdraw completely) civilian killings don't shoot up again.
If my understanding is correct, then the answer of whether it's working is inherently unknowable until after it's done.
JK
