<< GEL RULING IN ~ MAIN ~ HIX NIX CHICKS FLIX TIXS >>
CLUSTER BUSTER
Stephen E. Moore, who has some survey experience in Iraq, examines Lancet’s death estimate:
The key to the validity of cluster sampling is to use enough cluster points. In their 2006 report, “Mortality after the 2003 invasion of Iraq: a cross-sectional sample survey,” the Johns Hopkins team says it used 47 cluster points for their sample of 1,849 interviews. This is astonishing: I wouldn’t survey a junior high school, no less an entire country, using only 47 cluster points.
Read on. Meanwhile, Sydney academic David Burchell suggests Lancet boosters try a new approach: honesty.
UPDATE. Ha!
When you click on the first link “honesty”, check out the list of items on the right hand side of the page (at least on Thursday evening, Oz time).
“Stay the course with Blair”.
Posted by Stop Continental Drift! on 2006 10 19 at 05:49 AM • permalinkSo there are those who are taking this survey as gospel? And those who criticize are hacks? What the hell is wrong with these people? Do they honestly think that if this survey were true that people—that Iraqis!—wouldn’t have noticed? That we wouldn’t have heard about it long before now?
Never mind. I keep forgetting that reality, for some people, consists of their momentary whims and most fervent wishes. Bless ‘em should the actual, factual reality ever slaps them up side the head.
How long till Tim Lamb ert smears and libels this guy?
Posted by niobium2000 on 2006 10 19 at 06:20 AM • permalinkThe situation over at Tim Lambert is worth a look. TL did have some reasonable (I thought) comments about the WSJ article, but then Steven Moore himself actually showed up in the comments, very sportingly I thought, offering to answer questions.
But the whatever moderation system they have there apparently screwed up, and his post appeared out of order. Only one person seems to have noticed and addressed a question to him. Ad hominem attacks on him continued, perhaps by people unaware of his arrival on the blog.
I’m guessing Moore would be Bush’s main source of advice on survey issues in Iraq.
It’s actually all a strategic reframing. We’re not in Iraq to help Iraqis. If anything, they’re helping us.
A side effect is that they wind up with a representative government.
But Islamonutballism is our target. If that can be forgotten, then you can argue against the war in Iraq, they figure.
But nation building is not our hobby.
The jihadists follow our politics much more closely than people realize. A friend at the Pentagon just sent me a post by the “Global Islamic Media Front” carried by the jihadist Web site Ana al-Muslim on Aug. 11. It begins: “The people of jihad need to carry out a media war that is parallel to the military war and exert all possible efforts to wage it successfully. This is because we can observe the effect that the media have on nations to make them either support or reject an issue.”
Link here.Just as the Lancet’s timing has been the subject of comment, so is the upswing in attacks in Iraq suspicious, with attendant casualty figures feeding Kerry O’Brien’s attack on the PM tonight, as well as the ALP going for it in parliament.
Ecce Homo, as Iowahawk might sayOT those freedom loving czechs are at it again
The Interior Ministry has decided to dissolve the Communist Youth
Association. The reason given is that its programme statement says the
movement strives to remove the private ownership of means of production
and replace it with communal ownership, a ministry spokeswoman said.
The ministry says the statement is in breach of the Czech Constitution
and the Charter of Fundamental Rights and Freedoms. The association
plans to take the matter to court. The Czech counter-intelligence
service BIS lists the organisation among left-wing extremist groups.o/t ABC site headlines..
“John Howard denies he has weakened his position on Iraq”
and “Andrew Wilkie says John Howard sending troops to Iraq has increased the threat to Australia”.
drive time entertainment on ABC this evening….
end of P.M. radio national…David Hix and Major Moron.
Phatty Adams..The British RETALIATION to the Mau Mau mutilations and murders in Kenya. (40 or 50 years ago Phat).
Doing it a little too brown dear boy.Speaking of “clusters”,Bbc site has featured strongly “Lebanon mines continue to kill”,also “Rough justice (complete with torture pic from Abu Graeb closed down prison)Why are SOME countries MORE TOLERANT OF TORTURE..”
But hidden away on BBC site is this speech from NATO secretary general Jaap de Hoop Scheffer-on the subject of Afghanistan.
“Do not forget where the country came from.
We are there to defend our basic values,the basic values which guide U.K. society,Dutch society,Belgian society.
If we fail again then Afghanistan WILL COME TO US. It will become a breeding ground for terrorists again.
Do not forget 9/11 and other terrorist attacks.Do not forget the British terror attacks.”The blog linked to under the “Ha” above is appropriately represented by a picture of three chimps (I guess the theoretical room full of monkeys banging away at keyboards and finally coming up with grammatically correct sentences has become reality). And their apology for mistaking the two Stephen Moores is pretty much cut and run (with a cheap, Parthian shot at the Moore who did, in fact, author the WSJ article, calling him “a different hack named Stephen Moore”). The contributors are, I believe, denizens of the lumpen-professorial class, from which one has learned not to expect much in the way of deep thought, let alone fairness.
Lambert, perhaps, makes a reasonable point when he notes that Moore admits to a survey of 75 cluster points. Myself, I consider that difference of 28 points to be a minor issue; Moore also discusses the training and experience of the John Hopkins Iraqi field teams relative to other surveys here (point 4).
Moore is saying, “If you are going to do a small sample, try to do it right”, a point that whizzes over a lot of heads. Including Lambert, who sort of agrees when he uses the soup tasting analogy. Which is actually a valid comparison. Except that Lambert misses (as usual) the point that you stir the soup before tasting to make sure anything that settled on the bottom has a chance of being picked up by the spoon. Technique counts a great deal.
But Lambert’s reasonableness stops there.
Lambert responds to Moore here. In it, Lambert reverts to his usual bullying and diversion tactics. For example, Lambert harasses Moore about his survey experience, thus:
So how come you never thought to add a question about mortality to any of your surveys? How come these credible organisations have never surveyed mortality? It seems that JH folks are the only ones interested. How come?
Except that the JH folks aren’t the only ones interested (hello, the Iraqi Body Count? The United Nations?). And how does this “disconnect” disqualify Moore’s opinions? In Lambert’s eyes, because Moore disagrees with Lambert.
So, really, Lambert is not making a valid argument. He has a few good points, but his argument falls short because he assumes that the JH survey is valid, which may not be the case. I don’t think that it is, and the post-publication behavior of the JH team merely reinforces my opinion.
Posted by The_Real_JeffS on 2006 10 19 at 09:56 AM • permalinkIn the second link, I noticed (emphasis added)...
In 2002, according to a CBS-New York Times poll, a majority of US citizens believed that Saddam Hussein was “personally involved” in the September 11, 2001, attacks, presumably because the White House kept telling them so, without any reliable evidence whatsoever.
Some things never die.
Posted by Richard Cranium on 2006 10 19 at 10:20 AM • permalinkthe Iraq Body Count editors - who are themselves longstanding critics of the war - paid tribute to the brave researchers (mostly Iraqis) who ventured into dangerous neighbourhoods to collect mortality figures.
Now we’re beginning to get to the heart of the matter. Would these Iraqi researchers be the same Iraqis who provide stories and
fauxtographsphotographs to those intrepid journalists ensconced in hotels inside the Green Zone?OT:
Regarding the blog linked in the update (do I need to know those guys? Probably not):
I couldn’t help but notice that the guy didn’t actually bother to link to the OpinionJournal piece he felt was written by a “hack” (nevermind he didn’t even finger the right person…Stephen, Steven, no difference I guess), but merely to Tim Lambert’s treatment of it…is it just me or is that a favored tactic on the left side of the blogosphere these days, to incestuously link among several snark posts attacking the same subject, with only a few of them actually deigning to link to their target?
I don’t recall seeing much of that on righty blogs which generally both link an article/post and hat-tip the blog they’ve seen it mentioned on, but I’ve noticed the above behaviour on lefty blogs more than once in the past, and it seems to be a recurring pattern…anybody else see something like that, or am I totally out of my mind?
I didn’t notice that, PW, having read the OpinionJournal article a few days ago. But, now that you mention it, I have noticed that tactic being applied in other places, where I had to search for the subject article due to a lack of links.
Posted by The_Real_JeffS on 2006 10 19 at 01:05 PM • permalinkDuring a lull in follow-on operations planning at the Al’Qaeda Bureau of Shoe Bomb Affairs, Cave 134B, North Waziristan:
Achmed: My brothers, have you seen this? This man Moore has claimed it is necessary to count more Iraqi brothers than one must count of Christian pigs. Impudence!
Farouk: You refer to the fat pork-gobbling infidel Moore, producer of the important documentary regarding the crimes of Bush and the righteousness of the Muslim brothers?
Achmed: No, this is a different Moore ... Stephen.
Farouk: Steee-ffffen! Yech. It is like to spit phlegm. What kind of name is this?
Ali: A filthy Crusader name, you goat! So what does this dog say that we don’t already know? One Iraqi is worth many alcohol-swilling idol-worshippers. Especially Iraqis bravely martyred in marketplace carbomb operations.
Achmed: This is the problem. This Moore claims the correct estimate of 654,946 Iraqis martyred by vile Ameriki war crimes is not correct. How dare he! I am thinking fatwa ...
Ali: Wait, my brother. Maybe he has something. As I understand it, when they counted 1,841 Iraqi households, they found that more than 650,000 had been killed. So, if they count more households, then they will only find more martyrs. Especially if we put the word out to our brothers to speak the truth as we explain it to them. And you know, we need more martyrs. Just yesterday, when I was ushered into the presence of the Sheik, he said to me, why so few martyrs? More martyrs, please.
Farouk, lost in thought: You know, I’m thinking this shoe bomb is not such a good idea. The infidels know of this idea. The dirty bomb, the hijacking of the airplane. They know of all this. It is time for something new. Perhaps if we can humiliate them. I remember, when I was a little boy at the madrassa, how you two grabbed my shorts and pulled, thusly, so they became painfully embedded between the twin peaches of my buttocks. Al Wedgie. If only we could find a way to administer the Mother of all Wedgies ... hmmm ...
Ali: Stop your yammering, Farouk. We are discussing serious matters. It is like listening to the chatter of a Bengali’s trained monkey.
Achmed: Perhaps we should issue our own report, Ali? We take these martyrs the infidels have given us, and then, based on our best estimate of how many martyrs the Sheik wants, we multiply by maybe two, or three. Do you think 1.3 million martyrs is sufficient? Is 2.5 million martyrs better?
Ali: Do not insult the intelligence of the Sheik, unless you want to strap these martyrdom Nikes on your own feet! Clearly, we must at a minimum multiply the number of Iraqi martyrs by itself, to the power of the number of households questioned, in order to achieve anything approaching Islamic standards of accuracy. Meanwhile, for this Moore, I think you are correct, a fatwa is in order, as it is haram for infidels to count Muslims. Farouk, please run over to the Division of Fatwa Oversight in Cave 637D, and bring me a copy of Fatwa Requisition Form AQ001B. Be quick about it!
Farouk: Yes, I will just put my shoes on ...
Achmed: ... Not those ones, you donkey!
(Big orange cartoon blast and spiralling black smoke clouds, with unravelled dishtowels and Nike sneakers flying out of it. Smoke clears to reveal Achmed, Farouk and Ali blackened, with tweeting birds circling their heads. Enter a tall Arab.)
The Sheik: Salaam aleikum, my brothers. What have we here?
Achmed: ... It was Farouk, Your Excellency.
Ali: Apologies, Most Exalted, this unprincipled lump of camel ...
The Sheik: Farouk, my beloved, come to my arms! Martyring yourself and your friends before breakfast! My heart is filled with gladness. The courage and wisdom to test your Instruments of Righteousness on yourself! I have need of another lieutenant, a brave and resourceful man like you. Come with me.”
Ali: But ...
Achmed: He ...
The Sheik: You two have served me well. Your joyous moment has now come, to add your number to the ranks of martyrs. I would like you to put on these wonderful shoes you have designed, go to the West, and give it the old Madrassa try! What do you say? By this time next week, 72 virgins!
Walking away arm in arm with Farouk.
The Sheik: Let’s see, that’s 144 total. That’s a lot of virgins! I understand there is a bit of a shortage due to high demand, so maybe they share 72? After all, what is a virgin here or there between martyred brothers? ... Come, Farouk, I’m interested, do you have any thoughts on humiliating our Crusader foe?
Farouk looks back, grins and waves bye-bye to his crestfallen brothers.
Farouk: Well, Most Explosive One, it is funny that you should ask ...
Posted by crittenden on 2006 10 19 at 01:09 PM • permalinkJules
Given the unpremeditated nature of these “studies”, can you hazard an “estimate” on the results of the next one, perhaps dated (at a guess) sometime around Oct 08?Cheers
Posted by J.M. Heinrichs on 2006 10 19 at 02:38 PM • permalinkAnd a Canadian statistician-blogger rips the methodology, amongst other things:
http://tinyurl.com/y9cfwr
Also read the link at the bottom of the original post for another post on the Lancet “study”. (Found via the Relapsed Catholic blog.)Posted by andycanuck on 2006 10 19 at 03:46 PM • permalinkAs one who was quoted in the thread, and then misunderstood (possibly deliberately, given that the poster who misunderstood was even moderately literate), I’d like to clear up a few points.
One, cluster sampling is a methodology in places where we are unable (for various reasons) to use much better methodologies whose results are more dependable. I.e. a flawed tool is better than no tool. If you don’t believe me, look for cluster sampled research in industrialized nations. Cluster sampling has known problems.
Two, cluster sampling *is* used sometimes in cases of infectious diseases because its very flaws are actually a strength in such instances. Given that we’re ‘pretty sure’ an infectious disease is present in an area and we want to find out how many cases there probably are and how the spread, having what amounts to a non-random sample makes sense. That is, we want to sample precisely where we *know* the outcome of interest is *probably* present.
Three, I also questioned the Lancet study on the basis of not only the methodology but its application. That is, I contend that they sampled wrongly, and this seems to be confirmed by the detail Moore has found.
Four, I agree with using stratified sampling in this case, but simply stratifying by population is inadequate at best and deliberately misleading at worst. Stratifying the way the authors said they did all but guarantees that their (now unrepresentative samples) will be drawn from areas of very high violent activity. This is bias, built-in. As Moore says, if you’re going to do it, at least “do it right.”
Finally, not only are 50 clusters in a country of 27,000,000 inadequate (I would want at least 100, but 300 would be better), but given the sampling difficulties above (that I’ve discussed earlier) having only 50 all but guarantees outliers that bias the survey in the direction of counting too many deaths.
I’m sorry is some fail to see this as criticism of methodology, but that’s what it is. Sampling is decidedly contra-commonsense all to frequently.
Posted by JorgXMcKie on 2006 10 19 at 04:06 PM • permalinkRoberts responds to Moore’s WSJ article.
Apparently demographic data was collected. Moore was wrong. Sorry, this guy has no cred. for me any more.
“We’re not in Iraq to help Iraqis.”
Don’t tell the left wing loons. They seem to be laboring under the delusion that we’ve been bombing, blockading, and invading Iraq for the past 15 years in order to improve Iraqi living standards.
No point in waking them up.
Posted by Dave Surls on 2006 10 19 at 07:56 PM • permalinkApparently demographic data was collected. Moore was wrong. Sorry, this guy has no cred. for me any more.
dipole, when Roberts releases that survey data, and Moore is proven wrong, then credibility is lost. Right now, it’s at the “he said, he said” level. Roberts can easily prove that Moore is wrong (or lied) about the demographic data…..by releasing the survey data.
Roberts also said:
Let us address the discomfort of Mr. Moore and millions of other Americans, not by uninformed speculation about epidemiological techniques, but by having the press travel the country and tell us how people are dying in Iraq.
Isn’t the media doing that already? I thought they were—you’d think they’d be able to dig up confirmation for Lancet from past articles alone….especially given how the terrorists are given extraordinary access to their operations by western news agencies. Where are the stacks of bodies that are supposed to exist? This little inconsistency is Robert’s response is disturbing.
I simply can’t accept the word of Roberts at face value. When I see independent confirmation of the analysis as provided by the JH team, I’ll be ready to listen.
Posted by The_Real_JeffS on 2006 10 19 at 09:18 PM • permalink“Did anyone say cluster?”
It’s off topic, but…the more cluster bombs and WP dumped on the heads of Hezbollah…the better.
Posted by Dave Surls on 2006 10 19 at 09:36 PM • permalinkNo 28 The Real Jeffs notes that Roberts said “Let us address the discomfort of Mr. Moore and millions of other Americans, not by uninformed speculation about epidemiological techniques, but by having the press travel the country and tell us how people are dying in Iraq.”
Coming from a supposed scientist, that’s almost scandalous—he’s saying, if you don’t think my methodology is sound, well, go get your information elsewhere.
So much for peer review—Roberts clearly only likes THAT when it goes his way.
Posted by arrowhead ripper on 2006 10 19 at 10:58 PM • permalinkSurely you don’t think Roberts would lie about having the data?
Yes. Yes, I do think Roberts would lie about having the data.
Surely you don’t think Moore would lie about his documented* conversations with Roberts?
*: Remember? E-mails?
Posted by The_Real_JeffS on 2006 10 20 at 10:48 AM • permalinkHere’s another cluster fuck for you, Miranda:
http://tinyurl.com/yajus8Posted by andycanuck on 2006 10 20 at 01:11 PM • permalinkEh, anycanuck, Miranda is in permanent denial. Anything that threatens the integrity of his/her fantasy world is met with a resounding “LA LA LA LA LA!!! I CAN’T HEAR YOU!!! LA LA LA LA LA!!!”
But thanks for the link; the rational and sane portion of the world wants to hear the entire story.
Posted by The_Real_JeffS on 2006 10 20 at 01:21 PM • permalink
Page 1 of 1 pages
Members:
Login | Register
| Member List
Ha! Indeed.