Asymmetrical Information

Wikipedia also claims that Pepsi was always labeled "M" and Coke was always labeled "Q." I didn't know letter popularity matters so much, but there it is.

Very true. Some people think that "expert" = "automatically right", when they should be thinking "expert" = "should be capable of thoroughly documenting and explaining why his opinion is valid", and if that's not met, stop considering him an expert.

The Gladwell thing was a complete disgrace (to Gladwell). His entire response to Jane's questioning of the Ireland growth-from-demography was:

1) Two really sophisticated guys from a TOP university make COMPLEX, MATHEMATICAL model. (Note: he didn't explain what about it made it useful or accurate, or how it was able to rule out other causes. No, the mere fact that the guys went to a good college and made a "sophisticated, mathematical model" alone is a good reason to believe it, even if each and every step was wrong.)

2) After getting chummy with the researchers, I (Gladwell) was impressed by the study. (Yes, that's right, he's actually claiming that you should believe something *because* he was impressed with something a group of people got chummy with, did.)

3) Some gripe about how Jane wasn't respectful enough when attacking an "expert".

For example, it would have been nice if the people conducting the Lancet study had, while they were in the area, dropped by the local death-certificate-issuing-authority to see if their records confirmed the death rates they were getting

There is no such authority. Death certificates can be issued by a hospital, a morgue or a doctor. There are no such records; they are not being systematically kept. Even if there were, this would not give you an estimate of the death rate, because you would only get the numerator of the death rate (the number of deaths) and have no way of calculating the denominator (the size of the population from which that hospital's death certificates were drawn). This would, in fact, have been both impossible and pointless, which is probably why it wasn't done.

I have done a random sample of three of the statisticians quoted in newspaper reports as criticising the methodology of the survey, by the way, and so far the result is that 100% of statisticians were quite badly misquoted.

Les Roberts' response to your point:

"We have gone and looked at every recent war we can find, and only in Bosnia did all governmental statistics add up to even one-fifth of the true death toll. And in Bosnia, the rate was 30 or 40 percent, with huge support for surveillance activities from the UN. So it's normal in times of war that communications systems break down, systems for registering events break down.

"And in Saddam's last year of his reign, only about one-third of all deaths were captured at morgues and hospitals through the official government surveillance network. So, when things were good, if only a third of deaths were captured, what do you think it's like now?"

There is nothing wrong with countering experts, if you have something to base it on.

Both the NYT and the WP canvassed expert opinions and, in both cases, the majority of experts agreed the methodology was perfectly sound. The study itself was peer-reviewed and expert statisticians consulted for the project (and Daniel in comments indicates even some of the criticisms may be misquoted). You yourself have provided no methodological critiques. You did not claim the samples were wrong or misapplied or the calculations were off.

You offered two critiques. First, you claimed the numbers were too high based on other indicators. That’s not much of a critique since the other indicators are not very good. As Les Roberts, one of the authors, points out they looked at every recent war and government estimates are rarely even 20% of the actual numbers killed. So it’s not surprising that many of the available indicators are much smaller than the survey shows. In fact even when the Iraqi government was working it only captured about 1/3 of all deaths through their official counting systems.

Your second critique is that the errors are so big in studies like this they make the results useless. You are, in effect, impugning not just this study but all such studies. Quoting Les Roberts again the method they used is “the standard way of measuring mortality” in cases like this and the US government uses this method all the time. I guess they could all be wrong and you be right but I suspect it requires a little more effort on your part to show this.

As I commented on CT:

Why do so many people seem to believe that a gut-check is irrational? I.e., “this study is an order of magnitude greater than every other statistical source; are we really sure that extrapolating from a few hundred confirmed deaths is accurate?”

What’s wrong with asking questions like that? That’s the way we all reason when calculations turn out to have results that seem improbable. You’re balancing the checkbook, and you get to the end, and you have 10 times as much money in the account as you had imagined. Do you throw up your hands and say, “Whoo-hoo! That’s what the math says, and math is never wrong! I’m having a party!” Or do you recheck your work?

Or you're buying a few shirts at the department store, and the computer says that the bill is $10,000. Do you say, "Oh, well, computers are always right, so these shirts must be worth $10,000," or do you ask whether there was an error somewhere (i.e., a computer error, or the salesperson had a diamond ring on the counter that got mixed up with your stuff, etc.).

Of course, we don't *know* for sure that the Lancet results are wrong. But they do seem improbable given everything else that we know. So what's wrong with double-checking and thinking about whether the methodology was accurate? What's DD's angle here? It's not his study -- why is he so defensive about it?

You can perhaps accept that Lancet "did the best anyoune could do under the circumstances," but the only important question from a research point of view is: Is the resulting information useful?

And Robert's point is interesting, but it still begs the question. "So, when things were good, if only a third of deaths were captured, what do you think it's like now?" Well, what do I think it's like now, based on what? Because what I think is of little value, unless there is verifiable information supporting "what I think." Otherwise it's speculation, not a well researched study.

Patrick,

Yes, the information is useful. It's a well researched study using widely accepted methodology.

Yes, the information is useful (since it supports my previous opinion on the subject).

Somehow I doubt the same people would be defending a well researched study using widely accepted methodology that agreed with offically provided figures.

BruceR: do you have the source of that Les Roberts quote?

Hey, DD, you're a statistics guy, so:

When was the last time someone tried a study like this in a place like the US, where death statistics are pretty good? Do we know if it works under very controllable conditions? I'd feel better if this wasn't "widely accepted" based on the notion that it seems like a good idea but had actually been tested someplace other than Darfur.

Because I could easily imagine wandering a few urban neighborhoods and asking "Has any member of your household been arrested this year?" and concluding that half the population of the U.S. has had an annual run-in with the law. Or visiting a rural neighborhood and concluding that 80 million deer and 227 million ducks are killed by hunters each fall. Or hitting the suburbs and concluding that the US is 94% white, married, and college educated.

Obviously a survey designer or methodology who hopes to have a snowball's chance on the death rate in Iraq has to be able to come within a couple of percent in the U.S. So can they?

Daniel--so the researchers don't know what the denominator is for the local population? How are they extrapolating from their samples, then?

If they can't go to the hospitals, morgues, and doctors--and I find this hard to believe; there are not an unlimited number of hospitals, doctors, and morgues--they can go to the local graveyards/mosques, which presumably keep records. Since, according to various reports, these mosques require a death certificate for burial, and this is the reason that almost everyone who dies gets a death certificate, they should have a very good record of who in the area has died, particularly in urban areas where backyard burial is not an option. Obviously this measure is not perfect, but it would be a useful comparison, no? We can, after all, be reasonably sure that 90% of Iraqi dead are not being buried somewhere other than a graveyard. Indeed, the Lancet researchers could check this by asking the sample if the dead were buried in a graveyard or at home.

GT: Yes, the information is useful. It's a well researched study using widely accepted methodology.

GT, you continue to equate "accepted methodology" with "accurate". Come on! We all know that following accepted methodology (including peer review) only yields a greater probability that the results will be accurate. It's not a guarantee of accurate or useful information. It was not long ago that the studies, using the best methods available, and done by all the organizations involved with such work, ALL concluded Iraq had WMD stockpiles. It was certain knowledge. A slam dunk. Turned out the studies were wrong.

In this case, the Lancet study has a margin of error that makes it useless. If that means I'm slamming an entire industry that produces such reports, so be it. If a study told us that a particular beam needed to be somewhere between 2 inches and 10 feet in thickness in order to carry the weight of the bridge, we'd ask the authors to refine their numbers a bit. If a market study told us the foot traffic on a particular street corner averaged between 2,000 and 200,000 on any given weekday, it would be of no use in determining whether to build a store on that corner. The Lancet study tells us people have died as a result of the war. That's like telling us bears crap in the woods. We already knew that. The study does little to inform us about how many would have died under the sanctions (leaving Saddam in power) v. the current situation. That makes it useless.

David,

You make valid points but your last two sentences are wrong. The report explicitly compares mortality rates under Saddam vs the current situation. That's why they are called excess deaths, they are over and beyond what we would have expected based on the mortality rates of 2002.

GT:

The methodology is acccepted. It's a shame they didn't follow it. The authors admit, quite frankly, they they broke the randomness requirement by bypassing certain neighborhoods. They also (though they don't really mention it) break any sort of possibility of double-blindness for the study (or, in fact, single-blind).

There are too many chances in this case for someone to severely influence the results by steering the researchers into pre-chosen neighborhoods, or to skew the results in other fashions. Considering the admitted biases of the researchers, that constitutes a kiss of death for the whole study.

Sure cirby.

When was the last time someone tried a study like this in a place like the US, where death statistics are pretty good?

"Like this" in what sense? If you just mean "a cluster sampled study", then I think a lot of American economic statistics are in fact compiled in that way.

so the researchers don't know what the denominator is for the local population? How are they extrapolating from their samples, then?L

There are population statistics for each governorate. There are not population statistics for "the population who represent the catchmen area for this hospital". Even the population figures for the governorates are quite shaky and out of date and this is a source of uncertainty referred to in the study, but the "population" represented by a hospital is a completely undefined quantity.

If they can't go to the hospitals, morgues, and doctors--and I find this hard to believe; there are not an unlimited number of hospitals, doctors, and morgues--they can go to the local graveyards/mosques, which presumably keep records

You aren't half keen on presuming things, you know. Particularly, you're very keen on presuming that Iraqis love filing so much that they are prepared to keep the files up to date even in a state of war. I had an admin assistant like that once, but sadly she moved on, and it wasn't to Iraq.

but surely ... if they don't have accurate records of all the deaths, then Iraq might turn into a state of anarchy, where people were able to commit murder with impunity? Let me whisper ...

If a study told us that a particular beam needed to be somewhere between 2 inches and 10 feet in thickness in order to carry the weight of the bridge, we'd ask the authors to refine their numbers a bit

How about if there was an alternative consultancy down the road telling you that because the first report had so much uncertainty in its estimate, you might as well make the bridge out of tinfoil? Let's make the analogy complete here.

BruceR: never mind, found it.

GT: methinks you're not understanding the point, or just being wilfully obtuse. We're all well aware that the study purports to measure excess deaths. What the study doesn't do is give us a good idea of what that number is.

Jane and others have said, many, many times, that you can do the best possible study and still have a useless POS. Given the inherent problems in doing this study, and the very large range of the results, this is a useless study.

Professionals routinely can come up with well done work that provides no useful information. This study is sexy because it has a high headline number. Purely based on the variation of the study, it provides no information. When you get this kind of crap in the business world you do 2 things: fire the people who did the study or yell at them to tighten up their process and come back with a real answer.

While you can get published using this crap, and likely skate by for your Phd, there's no way that this would pass muster in an undergrad class! That's cause we actually check the work of undergrads for sloppy errors and ludicrous confidence intervals, while assuming that anyone at a high enough level is going to be honest and thorough enough to not have to worry about the basics like that. Far too frequently, that trust is misplaced and we get frauds like this.

As to dismissing the work of an entire field... um yeah. Epidemiologists have a VERY bad reputation for sloppy work, confusing correlation with causality, innumeracy, and the worst sorts of social science approach to research. This is mostly due to their backgrounds in the soft, squishy, quasi-sciences. They aren't quants and never will be quants. So trying to cow me into respect for a false authority... nice logical fallacy you've got there.

I do dearly love the idea that since government figures are unreliable, we must accept Lancet at face value. But the entire basis for the excess death figure is the previous government's statistics on deaths. So we're now denigrating the foundational number, dismissing the local authorities awareness of deaths (central government sure... but local authorities/service providers?), and claiming that our completely unverifiable number is the only one that could possibly be valid.

As we get deeper and deeper, y'all are sounding like Michael Bellisles and his misteriously destroyed, flooded records that were the only existing copies of key records. That's what's called "convenient". But if it is the new standard in authenticity and peer review, then I've got a desktop cold fusion power plant to sell you. You just can't look to closely at it or you'll interfere with the reaction... yeha that's it, interfere with the reaction!

Jane Galt:

"... But it nearly put the company out of business, just the same."

I doubt the Coke company was ever at any risk of going out of business.

There's something odd about Les Roberts' statement:

"We have gone and looked at every recent war we can find,
and only in Bosnia did all governmental statistics add up to even
one-fifth of the true death toll. And in Bosnia, the rate was
30 or 40 percent, with huge support for surveillance activities
from the UN. So it's normal in times of war that communications
systems break down, systems for registering events break down.

"And in Saddam's last year of his reign, only about one-third of
all deaths were captured at morgues and hospitals through the
official government surveillance network. So, when things were
good, if only a third of deaths were captured, what do you think
it's like now?"

It sounds totally reasonable, and in fact this is a priori
more or less what I would have expected the situation to be.
I would have been surprised if in the midst of war, the Iraqi
government had a really accurate count of how many people were
dying.

The problem is this is not what the study itself says.

Yes, I know, there is this assertion (page A5):

"Even with the death certificate system, only about one-third of
the deaths were captured by the government's surveillance system
in the years before the current war, according to informed sources
in Iraq. At a death rate of 5 per 1,000 per year, in a population
of 24 million, the government should have reported 120,000 deaths
annually. In 2002, the government documented less than 40,000 from
all sources. The ministry's numbers are not likely to be more
complete or accurate today."

Sounds reasonable. Except that Les Roberts' Iraqi coworkers went
and asked people for proof that the people they said had died had
actually died. And in 92% of the cases where asked the households
interviewed were able to produce official Iraqi government-issued death
certificates.

So here's the mystery. If the Iraqi government had a pile of
death certificates in 2002 that when added up would have amounted
to 120,000 or something similar, why then would they report 40,000?

And the same question for 2003, and 2004, and 2005, and 2006.
Except that by the time we get to 2005 the disparity between what
the Roberts study asserts and the Iraqi government claims has
gotten enormous.

Les Roberts and company's language is misleading. They say the deaths
weren't captured by the government surveillance system yet their
own data shows the Iraqi government surveillance system worked amazingly
well. No, what Roberts and company are really implying is that
there was a conspiracy to conceal the death rate. A conspiracy
that stretches all the way from before the invasion to the present.

No Mark, I think you are misreading.

From what I read there is no doubt that most people can get death certificates from hospitals, morgues, etc... People need those certificates for a bunch of things so they are diligent in getting them.

But the government is not good at counting those certificates. There is no centralized office that issues the certificates so they must be counted from the many places that issue them and it is that counting that has been found wanting. A hospital may issue a death certificate but, even prior to the war, 2 out of 3 of those were not counted by the central authorities.

The 655,000 figure sounds somewhat high to me, too.

On the other hand, what is the probability that the ultimate Iraqi death toll from this war will reach at least 655,000?

I'd guess 95%.

GT,

You're asserting by implication that there's a severe shortage
of people in Iraq that know how to add. Or that no one
in the government was interested in knowing how many people died
in 2002, 2003, 2004, 2005, or 2006.

Or that manpower shortages are so severe that the government
couldn't even find one person to count (because theoretically
one person could have done it) and note this is in country where
unemployment is a major problem.

Yes, one of these might be true but they all seemed highly
improbable. Therefore I didn't acknowledge them as possibilities.

Your last line is an assertion as fact what is really an assumption.

You're offering as evidence what I'm questioning.

You are saying that the Iraqi government didn't count deaths because
the Iraqi government didn't count deaths.

Here's another issue.

Reading a study like this (or any study) is an act of faith.
We implicitly believe that the authors are honest, rational,
and truth-seekers. If we don't believe that all these things
are true, or at least are willing to suspend disbelief that
they might not be, then there's no reason to pay attention
to such a paper at all because there are a thousand ways a
disingenuous and dishonest person might twist the 'facts'
from which people are trying to draw conclusions.

We can turn that idea around and do a thought experiment and
imagine that we are an honest, straightforward, truth-seeking
person like Les Roberts and ask what we would do if we were
in his shoes.

Now if I were such a truth-seeker, I would try not to
blind myself to data that contradicts my preferred conclusion,
nor would I try to hide from my audience problems with what
I prefer to believe.

Of course this an ideal. In the real world few people, few
scientists, come close to meeting the ideal. But some do.

How do Roberts and company rate on that scale?

The Iraqi death certificates data is a good test case. Where
it supports Roberts' conclusions it's up front and center.
Who could miss the claim that the overwhelming majority of
the claimed deaths were confirmed?

When the same data casts doubt on the studies conclusions
though suddenly the language becomes ambiguous and imprecise.
Suddenly there are no percentages to highlight the disparity
between what one set of data says and another. In fact
I suspect that 99% of the people who are not already aware
of the issue could read through the brief discussion
of it on page A5 without realizing the dimensions of the problem
at all.

So what does this mean? One reaction would be that they are
not impartial and therefore we shouldn't pay attention to
this study. I disagree. This situation is all to common
including among scientists. What it really means is that
this paper needs scepticism. If there is data that conflicts
with the author's conclusions then they are going to minimize
and obscure it and possibly not even mention it at all.

A more serious issue is why there is no dataset to go
with the paper. Maybe actually there is. I don't know
that there isn't but there's no mention of it in the paper
we're all reading. People don't need the names of the people
interviewed but they do need the responses and which cluster
they belonged too.

Now if I were an honest, truth-seeking researcher I would
produce this data in a moment.

DD,

What I meant was, has anyone tested the cluster-sampled study technique by trying to measure deaths in the US (or UK, or any other advanced Western democracy)? We have a good census, we have good recordkeeping at most county offices, and we have a fairly low rate of random unaccounted-for disappearance, so I'd expect that our official death counts would be pretty reliable. If the social scientists can make their estimates line up under these conditions, then I'll have some confidence that they know what their doing, at least in the "easy" case.

If they tried it and their estimates were way off, then there's a real problem with the method, however widely accepted it may be.

If, (as I kind of suspect), no one has ever even tried it, then we're left with a technique that looks good and is popular but hasn't actually been shown to work.

"You're asserting by implication that there's a severe shortage
of people in Iraq that know how to add. Or that no one
in the government was interested in knowing how many people died
in 2002, 2003, 2004, 2005, or 2006."

Gee, I can't imagine why the people in the Iraqi government would want to hide the fact that things are much worse than being reported. Surely they value poking America in the eye much more than they value keeping power.

It is not in the interest of anyone with the power and authority to count death certificates to actually do so. Is that really so hard to fathom?

There is no dataset with the paper because the authors have refused to release one. They did the same thing with their 2004 paper, despite repeated requests.

When scientists refuse to release their data, what inference do you draw?

So if the method is widely accepted then the conclusion is correct? Sounds like fun. Let me take a shot:

The sun rises once a day. The axis of the Earth is defined as north to south and is perpendicular to Earth's orbit around the sun; therefore the sun must rise in the east or the west. Therefore since there are two directions from which the sun can rise, every other day the sun will rise from the west.

Of course this is silly because I didn't include the direction of the earth's rotation. My point is that arguments consist of two components: form and premise. If the premises are flawed, no matter how perfect the form, the conclusion is not acceptable.

Boy, it seems I need a beer. TGIF

Didn't the Bureau of Labor Statistics just underestimate the number of jobs produced by the US economy last year by 800,000 or so. And no one was shooting at them.

From Mark's comment:
"If there is data that conflicts
with the author's conclusions then they are going to minimize
and obscure it and possibly not even mention it at all."

Any researcher who does this is a PR man, not a scientist.

From David:
"There is no dataset with the paper because the authors have refused to release one. They did the same thing with their 2004 paper, despite repeated requests.

When scientists refuse to release their data, what inference do you draw?"

If this is so, the paper is utterly useless. The very first step in checking the validity of a paper, is to analyze the raw data yourself. Even peer reviewers must do this to some extent, never mind anyone who undertakes a serious analysis.

There's absolutely no (scientific) point in discussing the paper's conclusions without access to the raw data.

Gee, I can't imagine why the people in the Iraqi government would want to hide the fact that things are much worse than being reported. Surely they value poking America in the eye much more than they value keeping power

Ah, I see we've gone from "the deaths are happening, they just haven't been recorded" to "the deaths are happening, and aren't being recorded due to a conspiracy".

Got no evidence? Invent a conspiracy that's hiding the evidence! Er, but what about the evidence for the conspiracy itself? Heck, let's just posit another conspiracy covering *that* one up. Someone call Oliver Stone and Kevin Costner, we've got a movie here with their name on it.

Or we could to the rational, scientific thing, and suspect that a study which "finds" hundreds of thousands of deaths undetectable by any means is not yielding good estimates.

GT,

I'd like to correct a mistake I made earlier. Although it's true
that I'm questioning whether the Iraqi government didn't count
deaths and the causal disregard of the official data in this
paper, on the other hand we do know that in 2002 the Iraqi government
didn't count deaths accurately. We know because 40,000 out of
a population of 24 million is an implausibly low number. Thus
if this survey suggests there were far more government death
certificates than the official statistics report then that would be
direct evidence that the government wasn't really counting the
certificates in 2002.

One explanation for would be conspiracy, a deliberate policy to
undercount, but it also occurs to me to wonder if a fee is or was
collected when these death certificates are issued. If so then it
could be the case that the people issuing the certificates are
pocketing the fees themselves and not reporting the deaths upstream.
That would be a more or less natural explanation of how this pattern
of errors could arise.

David Kane and Fred the Fourth,

I agree that if the authors of this study are refusing to release
their dataset, then:

"There's absolutely no (scientific) point in discussing the paper's
conclusions..."

The paper may look like plausible and it make look like science, but
in reality it's just a sophisticated fake if the dataset or dataset
absent real names is not released.

Further this observation has legs in that applies not only to the
paper, the authors of the paper and the journal publishing it, but
also to anyone approving of the paper. A "scientist" who knows that
the authors are refusing to release their dataset, and yet names it
science is a fraud himself.

I'm curious where the idea comes from that scientists normally share their datasets with anyone who asks. Maybe it is different in other social sciences, but in economics at least, this is certainly not the case. If a researcher goes to the immense effort of hand contructing an original dataset, he's not going to share it with other researchers (aka his competition) until he's milked it for every publication it's good for. Maybe after 10 years he'll start disseminating it.

I don't think this custom enables publication fraud very much - after all how hard would it be to make up a dataset in your computer that lines up with your claims? The only way of checking that your interviews are not fabricated would be to put other researchers in touch with your actual interview subjects.

Anne,

Do you really think there's any field where data is so cheap
and easy to get?

Is it really such a bad thing if someone else goes through
data you'd collected and notices a pattern you hadn't? The
rationale you're offering is a purely selfish one. It offers no
advantage to the community of economists as it slows down the rate
at which new understandings are gained. It practically guarantees that
very few eyes will look at a given piece of data.

It's bad news for the taxpayers who after all are likely the ones
financing these misadventures as it means far less bang for their bunks.

Why should we want this again? Am I missing something? Are there
any pluses to this behavior?

Shouldn't we all be trying to discourage it? And how would we go
about doing that exactly?

Wouldn't ostracizing people that do this sort of thing help?
Wouldn't refusing to publish papers that don't come with a public
dataset end it all pretty quickly?

As to detecting fraud I think you've answered the question yourself.
Any real world dataset contains far more information than is at
first glance obvious. There's a decent chance that someone making
stuff up will leave a signature of sorts behind because of all the
different things that person doesn't know about. The fabricator
is always vulnerable and never knows when their world might come
crashing down.

But even if there were no practical way to discover made up or
fudged data I hope I've made it clear that the scientific method
has more going for it than just improving the odds that fraud
will be detected.

Got no evidence?

Dan, there is plenty of evidence that the Iraqi government (and before them the CPA) censors mortality data. This has never been made a secret.

In medical research, data is usually shared with peer reviewers and with other researchers in the field, but not with the general public (a lot of the data is proprietary). In the case of the past Lancet study on Iraq, Les Roberts proved to be more than happy to send his data to anyone who a) asked politely b) showed some understanding of what to do with it and c) hadn't already proved themselves to be biased by making stupid and ill-informed hack criticisms. As a very low baseline, anyone who doesn't understand what a crude death rate is and goes around saying that the pre-war death rate estimate of 5.5/k is "much too low" really has no business trying to do anything with the raw material of this study.

Interesting allegation, dsquared.

What, precisely, do you think the term "crude death rate" is, and why do you think that half the European rate is appropriate for Iraq under Hussein?

Any researcher who does this is a PR man, not a scientist.

Not necessarily; it has been shown that even experienced professionals will completely miss evidence that falls outside of their training or expected scope of results.

In other words, it is possible for a researcher to minimize or ignore data that doesn't fit, entirely apart from willful intent. Peer review is one method of attempting to deal with this problem, but the efficacy of this somewhat depends on how much time the reviewership actually put into the review, and whether or not that reviewership might share some of the same baseline perspectives that lead to inadvertent minimizations or omissions in the authorship.

I still don't understand why people are not cheering this study. We were NICE (e.g. dropping concrete bombs to limit damage) to Iraq for three years and 655,000 people died.

Based on this study, anyone who argues we shouldn't invade Iran and North Korea is an idiot. Imagine the damage we could do if we were trying. Slam dunk.

DD,

I'll repeat my question one more time: has anyone tried a cluster-sampled survey death study like this one in the US or other advanced country where the official count is very likely to be close to correct? And if so, were there results close to the official count?

Put a different way, how do we know this "widely accepted" method actually works, even under ideal conditions?

You guys like evil conspiracy theories? I'll give you mine. Management was aware that New Coke wouldn't sell. The whole thing was a cover for a reformulation; Classic Coke is not the same as the original product. It contains far less of the decocainized coca leave, which contributed a wintergreen note to the flavor.

I have done a random sample of three of the statisticians quoted in newspaper reports as criticising the methodology of the survey, by the way, and so far the result is that 100% of statisticians were quite badly misquoted.

The problem is that Pepsi beats Coke, but New Coke Beats Pepsi, but old Coke beats New Coke.

Eventually Old Coke became New Coke after New Coke was off the market. Somewhat.

An interesting example, and proof that the real world is assymmetrical.

Twill00--

One would naturally expect to find a lower death rate in Iraq than in the US. The US has a relatively old population. In fact, at 5.5 per 100,0000, Iraq's prewar death rate was higher than that of most Arab countries. See this link for a more thorough analysis.

The study is comparing bad numbers with bad numbers.

"If 1/3 of deaths were being reported before..."
Holy-fucking-shit that's bad.

And I doubt it's worse now. Why would reporting be worse now than before? It should be better. There is little likelihood of deaths going unreported except in very rural areas, which isn't likely to impact the totals much. People want their losses recorded; if it can be done, it will be.

Mark:

And in 92% of the cases where asked the households
interviewed were able to produce official Iraqi government-issued death
certificates.

So here's the mystery. If the Iraqi government had a pile of
death certificates in 2002 that when added up would have amounted
to 120,000 or something similar, why then would they report 40,000?

The Iraqi government does not issue death certificates; doctors, hospitals and morgues issue them. As to whether they have a pile of uncounted death certificates, maybe they do... maybe they've been lost, or destroyed, or no one is being paid to count them.

Twill00:

Interesting allegation, dsquared.

What, precisely, do you think the term "crude death rate" is, and why do you think that half the European rate is appropriate for Iraq under Hussein?

Because

1) it matches the death rates of similar nations, and
2) Iraq is a young nation with a modern health care system. Older nations have higher death rates because older people die more frequently than younger ones.

There's nothing particularly foolish about having questions, but you could have found the answer to that question if your concern was finding the truth, rather than a reason to complain about the results.

Oh: For "crude death rate", you should be able to google the phrase and click "I feel lucky"; that got me to a Wikipedia article that had the term defined.

Comments are Closed.