Friday, October 13, 2006

Statistics, part II

Okay: a bit more on the statistics.

I've seen several people trying to wrap their minds around the numbers provided in the recent study published in Lancet, suggesting that there have been 650,000 excess deaths in Iraq since the war started. I'm going to address some of those issues here.

First, doesn't Iraq Body Count seem to provide evidence against the recent Lancet study?

No. Iraq Body Count tries to create a baseline minimum of "at least this many civilians have died, no questions asked". Their goal is to count

1) civilians only,
2) killed by violence,
3) reported by at least 2 English speaking sources.

Why do they have a range? I haven't investigated, but I imagine it's because there are times when an unknown number of victims are reported for an incident in the various source... plus, different primary sources might give different numbers. "Police believe 30 people were killed; observers report 50 people were killed; the hospitals are sure they've found indications of 33 victims", so you'd have a range of 30 to 50.

The Lancet study includes all deaths, for any reason, and would catch deaths that are not reported in the media at all (if the local strongman has just blown your friend's brains out, and says "and don't tell anyone about this, or you're next!", it's not going in the papers!), or are only reported by one source, or only by non-English speaking sources. Iraq Body Count provides a valuable service, but it's not even pretending to count the same numbers as this study.

Okay, then, doesn't this huge number of people killed prove that it couldn't happen? How could they handle that many bodies, and wouldn't people have noticed?

Well, it's a huge number, but it's only 2.5% of the population over 40 months. That's a bit over one half of one tenth of one percent of extra people killed each month. In the first few months, it would be unmitigated horror... but after that, it would numb to a kind of "isn't this how it's always been?" And remember that in those first few months, there was an active military invasion, with air, artillery, and cruise missiles strikes occurring frequently.

The additional deaths can certainly be handled by the number of people remaining (some bloggers have talked about "tripping over bodies", an obvious misunderstanding of the percentages), and, after the first year, it might not even have seemed all that unusual anymore. It certainly wouldn't seem newsworthy... "Lots of poeple die in war zone" isn't exactly unexpected, new, or startling.

What about the Iraqi government? Wouldn't they have a better handle on how many people have died, especially because the people surveyed had death certificates for about 80% of the deaths?

Well, death certificates are written locally; that means that they have to be sent to the central government. For over the first year, we were the government... and we'd just removed the old government. You can't get records flowing smoothly from the local government to the central government under those circumstances. After that, well, a lot of "collaborators" (i.e., Iraqis working with the occupying coalition forces) were targets of attacks. It would be very easy for records to get lost.

Finally, what kind of government wants to say "there were 650,000 deaths among our citizens where we only know about 50-100 thousand of them"? Who wants to admit that their citizens are dying at such an enormous rate? It's not just humiliating to admit to such powerlessness, it exposes a weakness to those who want to take power. Of course the Iraqi government is downplaying the study! They'd have to, whether they thought it was accurate or not!

So, neither Iraq Body Counts, Iraq government records, nor the lack of reporting makes these numbers unlikely.

They are reasons for people to be suspicious, to make them want to dig for the truth... but they are not reasons to try to ignore these numbers. Yes, they are horrible. That's all the more reason to dig in and make sure we know the truth, because as horrible as the numbers are, it would be that much more horrible to look away from them if they were true.

Comments:
I agree. "Not liking the number" is not a legitimate reason to dispute the data.

However, there are suggestions that the "95% confidence level" of the study is way too high, and that the sample size was still too small to justify the extrapolation.
 
I'm afraid that an "interval" can't exactly be "too high". Do you mean "too wide"?

The sample size is by no means too small to justify the extrapolation. To say so demonstrates a profound misunderstanding of statistics and sampling.

If you respond with more appropriate comments, I'll be glad to admit that you might understand what you're talking about... but right now, if you *do* know what you're talking about, you know exactly why I'm objecting, and understand my unwillingness to be more kindly here.

If you don't understand why I'm objecting, then you're talking about things you don't know about, without qualifying your statements... which, in my book, is dishonesty.
 
I tried to reply to you via the Liberal Avenger but apparently, intelligent conservatives upset their delicate constitution as they refuse to post my response. I would say that Avenger is rather a misnomer, wouldn't you?

I wanted to point out that Iraq Body Count has done an analysis of the Lancet article and they have concluded, ``In the light of such extreme and improbable implications, a rational alternative conclusion to be considered is that the authors have drawn conclusions from unrepresentative data. In addition, totals of the magnitude generated by this study are unnecessary to brand the invasion and occupation of Iraq a human and strategic tragedy." This is exactly what I was saying--death is either a tragedy or it's not but desperately clinging to a ridiculous figure to prove a point is dishonesty.

Since their political leanings are similar to yours, feel free to refute their analysis.
 
Post a Comment

Links to this post:

Create a Link



<< Home

This page is powered by Blogger. Isn't yours?

Weblog Commenting and Trackback by HaloScan.com