March 22, 2005

smijer: bad liberal

from - smijer

This view of depleted uranium weapons is popular among us lefties. After a fair amount of reflection, I've come to believe that this view is too radical, and not scientifically sound.

The World Health Organization has compiled a fact sheet which brings the dangers of depleted uranium weapons into better scientific perspective. We do need awareness, and we do need to hold our government accountable for providing the correct clean-up and isolation measures if they are to continue to use DU instead of safer alternatives. However, I feel we harm our case when we overstate the case. It is far too easy to dismiss an issue entirely when one feels that one has been lied to about it. That's the danger of exaggerating the dangers and the health impact as it appears some lefties have done on DU.

So, I've parted ways with the most outspoken liberals on ANWR and depleted uranium. I guess that makes me a bad liberal.

P.S. Ah, hell. While I'm at it, I should go ahead and call for other libs and anti-war types to refrain from using the 100,000 civilian casualties figure. When we're called out for hyping artificially inflated numbers, it will detract attention from the fact that we have certainly killed more innocent Iraqis than Americans were killed on 9/11 by a factor of at least two or three.

::

Posted by smijer at March 22, 2005 06:59 AM
Comments

Before you rush to labeling yourself a "bad" liberal, try doing some more research--and don't just look at sources that are "official." Take a look, for instance, at Leuren Moret's article in the San Francisco Bay View last week (see http://www.sfbayview.com/081804/Depleteduranium081804.shtml) or Carol Sterritt's article in a Marin County, California, newspaper (see http://www.coastalpost.com/04/08/01.htm).

univar.jpg Posted by Daily Bailout on March 22, 2005 07:22 PM
Link to comment

Your brushing aside of the facts of Iraqi deaths needs some caution. BE CAREFUL TRYING TO COME TO CASUAL CONCLUSIONS!

As far the 100,000 Iraqi civilians killed, that number could be higher, as some researchers estimated the toll at 185,000. But remember, the study of those civilian deaths--Iraqi men, women, and children--by Lancet was finished last summer. So those numbers are still climbing.

Considering your risky jump of faith in mainstream media, you may want to pick up the numerous books now available on our CENSORED press. And you should read the ACTION ALERT by FAIRNESS & ACCURACY IN REPORTING at http://www.fair.org/index.php?page=2472.

univar.jpg Posted by Daily Bailout on March 22, 2005 07:59 PM
Link to comment

Well DB, first let me say that I didn't rush to either of these views. Second, I'm not relying primarily on the media - either "official", "mainstream", or "alternative".

Months ago, I got interested in the depleted uranium claims, after watching a flash movie from Eric Blumenthal about it. I decided at that point that I was going to raise a very big fuss about it (on this blog) if the facts were as they were portrayed in that movie. Its a habit of mine to check my facts before I start raising a fuss. So, I looked at the reports from people like the WHO - not because they were "official", but because they were "scientific". The WHO is not a partisan organization, and they have no motive for white-washing the impact of DU - they are an organization devoted to, above all else, improving the quality of health around the world. They employ a scientific methodology, where cherry-picking data is not allowed, and the scientific import of the data must be fully considered: methods that no media outlet or advocacy organization are bound by.

You'll see that the WHO reports are not a whitewash. They do recognize limited dangers from DU, and they do recommend certain procedures to guard against the dangers of contamination.

The WHO reports are not isolated. This commentary from a U of MD school of medicine professor refers to quite a large body of research on the effects of DU that supports conclusions similar to the WHO.

With respect, I read the articles you included in your comment, and I saw no safe-guards against data-mining (selective use of data to support a conclusion) and other scientifically unsound practices.

On the Lancet study, I was aware of it from a relatively uncritical portrayal of it in the mainstream American and British media just prior to the election. At the time, I was suspicious of it because it was an outlier... because it returned results so dissimilar to the other estimates gathered by various methodologies. Furthermore, since that time, there have been some flaws exposed in the Lancet methodology. After looking at that, and comparing their result with other results, I've come to the conclusion that 100,000 is an irresponsibly high estimate of Iraqi casualties.

Whether its 10,000 or 100,000, Iraqi casualties, it is far too many. I think we can agree on that.

univar.jpg Posted by smijer on March 23, 2005 07:05 AM
Link to comment

Let's address your final point first--namely your insensitivity to numerical value.

Based on your logic, it does not matter whether you earned $10,000 or $100,000 per year. The amount does not matter in your world view. This is just a matter of degree.

Fortunatley most individuals, families, businesses, and governments (not our own) do accounting. By contrast, your casual approach to numerical significance seems incredibly cavalier and silly.

Of course, 10,000 or 100,000 or 185,000 or even 10 civilian deaths are too many. And that's the point here--the higher number, the greater the toll on Iraqi civilians, a cost if high enough will almost certainly come back to haunt America.

So, yes, the numbers do MATTER. In fact, that's why you lowered the number of killings to begin with. This implicit contradiction cannot stand without weakening your argument.

However, if, as you say, only 10,000 Iraqi civilians have been killed because of the war, then you have estimated that there was only one civilian death per every 20-square-mile area in Iraq. I, on the other hand, implore you to consider how much more reasonable it would be to put the civilian death toll at one death per square mile in Iraq. This sampling not only seems more realistic, but perhaps understated.

univar.jpg Posted by Daily Bailout on March 23, 2005 05:25 PM
Link to comment
Let's address your final point first--namely your insensitivity to numerical value.

I really don't understand why you allege that I am insensitive to numerical value.

Based on your logic, it does not matter whether you earned $10,000 or $100,000 per year. The amount does not matter in your world view. This is just a matter of degree.

I don't think you understand my worldview. I really don't. Maybe if you said that according to my world view, we could both agree that neither $10,000 or $100,000 would be enough money for me in a year, then you would be closer.

Of course, 10,000 or 100,000 or 185,000 or even 10 civilian deaths are too many.

Yes, this is something we can agree on. That's why I ended my comment with a similar sentiment. I like to be agreeable.

In fact, that's why you lowered the number of killings to begin with.

I do wish I had the power to lower the number of killings. In fact, I merely rejected the higher estimates as an irresponsible claim, not supported by the evidence.

However, if, as you say, only 10,000 Iraqi civilians have been killed because of the war, then you have estimated that there was only one civilian death per every 20-square-mile area in Iraq.

Considering how much of Iraq is uninhabited desert, and considering how little I know about the geographical distribution of violent engagement there, it would be very irresponsible for me to guess about how many civilian deaths per 20-square-mile area had taken place in Iraq.

I personally don't have an estimate that I feel confident in. I feel that the lower boundary is pretty well defined by the consensus of estimates in the area of 5-15, mostly in the 10,000 range, thousand from a variety of sources. So, when I feel it necessary to speak about a numerical estimate, I tend to use minimum terms, so as to avoid ruling out the possibility that the actual number is higher... even possibly much higher. I reject the estimate from the Lancet study because it is statistically useless. Read the Slate piece to see why, and read Instapundit to see why Kaplan's piece is correct in this regard. I don't agree ideologically with Glenn Reynolds, but the facts are pretty plain and he conveys them accurately.

Accuracy about the numbers is important, and we are not well served by hyping a figure that is not supported by the data.

univar.jpg Posted by smijer on March 23, 2005 06:14 PM
Link to comment

Here I am up at god knows what hour finishing some work. And I of course check in with you.

You did not address the critical contradiction in your argument: namely, a call for accurate numbers versus a regret for too many deaths.

I can understand why you want to find a point of agreement (e.g. we both feel that any loss of civilian life is disturbing), but to find that point we must both agree on a number first. And here's why: Because you start by discounting a complex study based primarily on the lame rebuff and false analogies of Kaplan at Slate, and because you have made numerical value so important to your argument. In a nutshell, either exactness counts or it doesn't. That's my complaint with your argument. You can't have it both ways without losing my interest in taking your argument seriously.

univar.jpg Posted by Daily Bailout on March 24, 2005 01:54 AM
Link to comment

DB... once again... exactness counts to the point where it can be achieved. Inaccuracy is to be avoided. One doesn't gain anything by adopting an exact number when the exact number isn't justified by the data.

I read the Kaplan piece, and other pieces critical of and supportive of the Kaplan piece. To me, Kaplan and his supporters are far more convincing than those critical of him. Unfortunately, the Lancet study requires a paid subscription, so I cannot read it directly... I only have press accounts of the methodology and results. However, provided that Kaplan is correctly relaying the methodology and results reported in the study, his conclusions about the 98,000 figure are sound. I would challenge you to demonstrate why you think that his rebuff is "lame".

If we are to appeal to the results of the study, despite its difficulties with samples, then we could at least report it as a confidence interval... as that was what the study provided. We may not score rhetorical points by saying that we are 95% certain that the number of Iraqi civilians killed is between 8,000 and 194,000, but putting it that way will at least be an accurate representation of the study's result.

The simple fact is that, desirable as it is, we do not have a number that we can confidently call an accurate estimate of Iraqi civilian casualties. The best we can do, with reasonable confidence, is establish a lower threshhold and say that the number is not likely to be much below it. IBC's work does a pretty good job of that.

If you feel that we have to agree on a confident estimate before we can agree that too many have died, then I'm afraid we will never be in agreement. I feel that too many have died even though I'm not sure whether it is 8,000, 15,000, or more. If you disagree with me on that, all I can tell you is that, sadly, you aren't the only one.

univar.jpg Posted by smijer on March 24, 2005 10:41 AM
Link to comment

smijer, your words are golden but your numbers are still fuzzy.

You, of course, admit that you can't reasonably come up with an accurate number. So, I ask you, why settle on any number at all?

Settling on an exact number, or a probable estimate, is the key to understanding what the costs of this war have been. And, as you prefer, a number that will ring in the ears of listeners without the audience attacking us as overblown liberals would be good too. But before we settle on a figure, we have to decide what is a "good" number. And here I (a liberal to the hilt) must take my liberal friends, including you, to examine the calculator of our message.

Should we define a "good" number as that which best fits into how our megamedia networks circumscribe political debate and issues, so that we sound, at least, somewhat mainstream?

Or should we challenge these media-message shapers when they are up to no good, or are simply wrong, despite our Sysiphean task at hand?

I.e. we liberals cannot shy away from the nitty-gritty dirty-hands work of digging into the math of statistical science or analysis of any other difficult study or political terrain. And we must trust the expertise of the scientist rather than the rhetorical reactions of the right.

And, yes, you did hear me correctly. I said the megamedia are wrong in this case. And so begins my apologia.

To start with you must (please) read Daniel Davies if you have not at Lancet Roundup. One among other astute critics of those true believers in the Bush Administration and the Establishment who don't know what to do with science and logic except to dimiss it immediately, Davies should (I am confident) set the record straight on this matter.

So until you have seen his scathing, intelligent and logical dismissal of Kaplan et al, I will calmly await the results of your new study.

univar.jpg Posted by Daily Bailout on March 24, 2005 12:54 PM
Link to comment

Is it my numbers that are fuzzy? How about this... why don't you tell me what percentage confidence we should have in the figure of 98,000 total dead. Can you show me, mathematically, how you arrived at that level of confidence? If you can be, as the paper states, 95% confident that the number of "extra" deaths is between 8,000 and 194,000, then how confident should you be that the correct number is between 90,000 and 106,000, for instance?

On the separate issue of how the sample was prepared, have you an explanation for the apparent undersampling of northern and southern Iraq?

univar.jpg Posted by smijer on March 24, 2005 03:34 PM
Link to comment

First and foremost, and this is where we differ the greatest I can see, I put my trust in the expertise of the statisticians and epidemiologists who did this study, not the media network boobs who revised the study's results because they either could not imagine the number of Iraqi civilian deaths at or above 100,000, or did not understand the statistical principles undergirding the study.


It befuddles me that so many good meaning liberals fall for this kind of revisioning, yet all the while countering the right wing for perpetually trying to revise America's founding and constitution to read like the setting up of a theocracy. And why do we attack the right wing revisionists? Because the historians have done their work. History, done well, will answer the fundamentalist conservative. Likewise, in the case of the Lancet study, statisticians and epidemiologists have done their work. And let's first try to understand what they are saying.

Davies clearly deconstructs the "Kaplan Fallacy" by showing the use of "asymmetric rhetoric about a symmetric confidence interval." In other words, Kaplan talks only about how the numbers should be less, not more, when in fact the numbers could be at about 195,000. Indeed, the danger in all textbook cases of cluster sampling is that the researchers will underestimate, not overestimate the numbers. Given that expectation, the confidence interval is often wide.

I like Lila Guterman's Counter to Kaplan in Columbia Journalism Review: But he was wrong. I called about ten biostatisticians and mortality experts. Not one of them took issue with the study’s methods or its conclusions. If anything, the scientists told me, the authors had been cautious in their estimates. With a quick call to a statistician, reporters would have found that the probability forms a bell curve — the likelihood is very small that the number of deaths fell at either extreme of the range. It was very likely to fall near the middle.

What is very interesting is how many journalists and bloggers critiqued the study without even reading it. Andy S does this very thing at, ironically called, The Age of Unreason blog.

That is why Tim Lambert at Deltoid comes down hard on those critics who either did not read the study or have no idea the first thing about statistics. Here's Lambert in his own words: I don’t want to seem harsh here, but if you haven’t even studied basic statistics and you criticize the statistics of a study that has been peer reviewed by professional statisticians you are likely to end up looking pretty silly.

For an intelligent conservative point of view, namely from The Economist, see Counting the casualties. I like how the writers handled the Fallujah question in the study: This concern [the cluster representation of violence] is highlighted by the case of one cluster which, as the luck of the draw had it, ended up being in the war-torn city of Fallujah. This cluster had many more deaths, and many more violent deaths, than any of the others. For this reason, the researchers omitted it from their analysis—the estimate of 98,000 was made without including the Fallujah data. If it had been included, that estimate would have been significantly higher. Again, bolstering the argument that not only were the professional researchers wise and cautious, but the numbers of dead could really be much much higher.

univar.jpg Posted by Daily Bailout on March 25, 2005 06:09 AM
Link to comment
First and foremost, and this is where we differ the greatest I can see, I put my trust in the expertise of the statisticians and epidemiologists who did this study

Well, this is a major difference. One study does not a scientific consensus make, and even the very best researchers often make mistakes. If there were a scientific consensus, based on multiple careful studies, around a 100,000 "extra" deaths figure, I would be more accepting of an appeal to authority on that basis.

not the media network boobs who revised the study's results because they either could not imagine the number of Iraqi civilian deaths at or above 100,000, or did not understand the statistical principles undergirding the study.

This is another misunderstanding. My critique is less of the Lancet study itself, and more of the media's and many pundit's often misleading representations of it. Remember, the study did not produce a number - at least not one with an associated statistical confidence - it produced an interval... just like in the political polls. The confidence interval is expressed as 8,000-194,000 but it could just as easily have been expressed as 101,000 +/- 93,000. The problem is when that 101,000 is taken out of the context of that +/- 93,000. With the +/- 93,000, you see just how ridiculously uncertain 101,000 really is. Yet, what do we hear in the media, and from some anti-war pundits? We hear headlines about 100,000 deaths.

Davies clearly deconstructs the "Kaplan Fallacy" by showing the use of "asymmetric rhetoric about a symmetric confidence interval." In other words, Kaplan talks only about how the numbers should be less, not more, when in fact the numbers could be at about 195,000.

Kaplan's headline is clearly fallacious, and Davies criticism is accurate. Kaplan's point remains strong. There is no certainty around a 100,000 figure.

Indeed, the danger in all textbook cases of cluster sampling is that the researchers will underestimate, not overestimate the numbers. Given that expectation, the confidence interval is often wide.

And it would be legitimate to ask and study over the question of how close the Lancet study is to a textbook case. The methodology sampled towns of greater population more frequently than lower population towns (probably a large part of the explanation for why central Iraq was oversampled, leaving areas where the death rate apparently went down relatively undersampled). If there was more violence and more infrastructure damage in larger population centers, then that would result in higher estimate than actual.

However, and I can't stress this enough, all of this is just second-guessing. Unfortunately, it was impossible for the researchers to do a standard survey, and therefore we are left with uncertainty. We cannot use logic to extract more precision from the study than was built into the study by the available methods.

What is very interesting is how many journalists and bloggers critiqued the study without even reading it. Andy S does this very thing at, ironically called, The Age of Unreason blog.

Actually, he did a preliminary critique without reading it, then upon reading it posted an update. But once again, the problem is not so much with the study, but with how it is presented to the public.

I don’t want to seem harsh here, but if you haven’t even studied basic statistics and you criticize the statistics of a study that has been peer reviewed by professional statisticians you are likely to end up looking pretty silly. -[Tim Lambert]

That's certainly true. However, many of the critiques have been done by people who are pretty well aware of the process. Pre-publication peer review is a wonderful thing, but it is generally just the first step in the process of validation - not the last.

For an intelligent conservative point of view, namely from The Economist, see Counting the casualties. I like how the writers handled the Fallujah question in the study

Indeed, the economist piece is intelligent. We shouldn't kid ourselves.. the economist has been trending more our way in the last couple of years, partly because of their objections to the war. Nevertheless, apart from their selection of headlines, they did a good job covering many of the relevant points about the Lancet study and reporting about it. They were sure to point out in no uncertain terms that:

Statistically, 33 is a relatively small sample (though it is the best that could be obtained by a small number of investigators in a country at war). That is the reason for the large range around the central value of 98,000, and is one reason why that figure might be wrong. (Though if this is the case, the true value is as likely to be larger than 98,000 as it is to be smaller.)

Equally important is this statement:

The way forward is to duplicate the Lancet study independently, and at a larger scale.

Another good piece comes from the Washington Post, which includes this:

"The methods that they used are certainly prone to inflation due to overcounting," said Marc E. Garlasco, senior military analyst for Human Rights Watch, which investigated the number of civilian deaths that occurred during the invasion. "These numbers seem to be inflated."

A senior analyst for Human Rights Watch is not someone I would classify as a "network media boob".

While we're on the subject, I should point out one other fact, I haven't seen addressed much elsewhere. Since this study relies on calculations of death rates, taking the baseline number of deaths from applying pre-war death rates to the time period between the beginning of the war and the dates of the study, it is subject to change in more than one direction as time goes on. IF the death rate today, or in the future, is lower than the pre-war death-rate, then the number of excess deaths they compute will begin to go down, and possibly even disappear. So another criticism of the Lancet study is that it only represents a snapshot of a specific time, and it is impossible to guess what today's figure is - even as a minimum - from what last year's figures were. I would guess that the deathrate has begun to decrease, but I would guess that it is still at least a few percentage points higher than pre-war death rates. But again, this is all guesswork.

Again, if we are going to rely on this single study - warts and all - we should at least be honest about how we present it - warts and all. Instead of pronouncing 100,000 dead, we should talk about 100,000 plus or minus 90,000 dead. I personally find it more cautious and sensible to take the numbers that are certainly dead, and add the words "at least" in front.

univar.jpg Posted by smijer on March 25, 2005 10:47 AM
Link to comment

smijer, one of the best responses I've seen in years by a liberal. Thank you. But I still a bit more to add. I will follow up tonight. I'm taking my son out right now. JB

univar.jpg Posted by Daily Bailout on March 25, 2005 11:11 AM
Link to comment

You've certainly made it difficult to convince yourself. Again I must add that if you cannot have any confidence about 100,000--that is because studies like this one tend to err on the side of underestimating. So the best guess would be more than 100,000.

univar.jpg Posted by Daily Bailout on March 26, 2005 01:41 AM
Link to comment

Finally, I have a moment to respond intelligently. And I am delighted as well. For I have lots to address. Because your argument hinges on three key appeals--your appeals to authority, to attacks of statistical sampling, and to the post hoc fallacy--it will be hard for me to address all three without a thousand or more words. So let me start with your fragile appeal to authority in this response, and take up the other two in following responses.

Authority: A very important appeal in your argument. Marc Garlasco is critical to your argument, and to be fair you have obviously quoted him from a reputable newspaper. And you're right, he's not a media network boob. But, sadly, he was the VICTIM OF MEDIA NETWORK BOOBS--as admits he himself.

Although a senior military analyst at Human Rights Watch, even such experts make mistakes. In fact, he has recanted since reacting initially to the media. He now says that his quote in the Washington Post was "really unfortunate" especially since he had not read the Lancet paper at the time (that's really not the best way for an expert to react, btw). "Like any good journalist, he got me to comment," regrets Garlasco. The analyst told the Post reporter, "I haven't read it. I haven't seen it. I don't know anything about it, so I shouldn't comment on it."

Victim of American media charlatanism, Garlasco was suckered. He told the Chronicle of Higher Education that he "misunderstood the reporter's description of the paper's results . . . did not understand that the paper's estimate includes deaths caused not only directly by violence but also by its offshoots: chaos leading to lack of sanitation and medical care."

Garlasco is a critical case in this larger debate, not just because it weighs in for the strength or weakness of your argument, but also because Garlasco's professionalism and credibility have been put on the line. In the real world of professionals, a history of expert testimony can make or break a career. Professionalism demands the kind of integrity which the mainstream American media pretends to respect but so often slouches from.

It is clear that the overuse of Garlasco's mistatements (not his afterthoughts) has been fodder for those who want to discredit the Lancet study for some particular reason that we have not yet gotten to.

univar.jpg Posted by Daily Bailout on March 27, 2005 06:05 PM
Link to comment
Comments for this entry are closed. Please leave your notes on a more recent comment thread.