« Ladies and Gentlemen, Your Friend and Mine, Michael Ledeen | Main | Electric Light Still Struck Like Arrows »

July 05, 2007

Comments

Not a major complaint, but perhaps the phrase "General Petraeus has two whole months and part of September to show a Democratic Congress that the current operation is worth supporting and funding." should have the word Democratic dropped out since the number of Republicans who are publicly displaying major reservations is increasing.

And the 80-90 " al Qaeda wannabes" crossing over is not something to make one shudder. And one has to question how much control Syria would have over that anyway.

As far as Iran goes, it seems there is little we can do that wouldn't make things worse, specially since Iran has its own domestic issues at the present time. Any overly strong movement on our part would quite possibly make Iran more difficult to work with.

Unfortunately, you also pointed out that the area where the most progress is needed, the ploitical front in Iraq, there is little to no progress. Without that, nothing else matters in the long run.

Looks good to me. Next spring, when the surge is complete and the Iraqis have stood up, let's ramp down the occupation to 40k-50K soldiers and marines tucked safely in the coalition's longterm (but not permanent!) bases on the outskirts of town.

Who's with me?

Where are the statistics from? I looked at the links but couldn't see.

So we've got more than forty civilians being killed every day in Iraq, basically no progress with governmental reform, and a near-record high in our own military casualties, and that's a success? Another such victory, Pyrrhus, and we are undone.

Where are the statistics from? I looked at the links but couldn't see.

He gets his civilain casualty numbers from here, there's a wiki that shows suicide bombings in Iraq, and I presume that U.S. military casualties comes from MNF-Iraq.

Every year there's a dip in insurgent opstempo at some point before it ratchets back up through late summer and fall. We'll need to see several months of flat to declining attack numbers before there's any reason to read June as a consequence of the Dribble. See the Iraq Index at Brook.edu, particularly the graph of average daily insurgent attacks since 2003 to have the general ratcheting pattern smack you in the face.

There is another view of this, almost an opposite view, over at www.phoenixwoman.wordpress.com (Mercury Rising), that all of you should look at. The figures they quote indicate a huge increase in the number of bodies being dumped in Baghdad and elsewhere. Has Charles any comment on this apparent contradiction?

I don't see much evidence of progress in these charts at all. The fact that June had better numbers than April or May might mean progress, or it might mean that April/May was simply an aberration and we're now regressing to the norm.

Mind you, there's been plenty of expectations management coming from the highest levels of our government. The administration has repeatedly warned everyone to expect a violent summer, as the enemy supposedly tries to influence our decision to stay. Of course, this is a can't-lose prediction, since either things get better which is great, or things get worse which is exactly what we predicted.

I'm tired of playing this game. When things get worse, no one admits that we should be more inclined to leave Iraq. So if things get better, why am I expected to view it as evidence that we should stay?

Charles, how do you reconcile the above with the report in the Washington Post that bodies found on the streets of Baghdad are up 41% from January?

http://www.washingtonpost.com/wp-dyn/content/article/2007/07/04/AR2007070401744.html?hpid=topnews

And "casulties are down" in relation to what? It's still higher than last October, which was still more than 600. Considering the fluctuations we've seen, I'd be very very cautious before claiming this is more than a statistical blip. I'd like to see six months of an obvious downward trend before feeling optimistic.

I'm afraid I'm not willing to say there is anything more than statistical noise on the charts presented. Single month trends have been reversed too often in Iraq.

"I'm afraid I'm not willing to say there is anything more than statistical noise on the charts presented. Single month trends have been reversed too often in Iraq."

Total agreement. I suspect 3 month trends compared to other 3 months would be somewhat better, but still too much noise.

in addition, there is the fact that the US military officials are refusing to measure or track civilian casualties. If they wanted to provide us with solid numbers, they could do a lot more. Instead, we get wobbly numbers.

Hooray, the wobbles wobbled down today!

I'm with Seb.

Moreover, the stated goal of the surge was (quite sensibly) not just to bring casualties and violence down period. ("Sensibly" not because this isn't a good thing to do, but because the likelihood that a troop surge of the sort we are able to provide, over the period we're able to sustain it, would be able to produce a serious and lasting reduction is basically nil.)

It was to reduce violence in order to provide a breathing spell so that the Iraqi government could make political progress, thereby bringing about some sort of reconciliation. With enough political progress, the surge would have given the Iraqis the chance to pull themselves out of their tailspin. Without it, it's just a way of tamping down the violence temporarily at a significant cost to our own troops, while postponing the question what to do.

And, of course, there has not been much (if any) political progress. Nor was there much reason to think that there would be.

...how do you reconcile the above with the report in the Washington Post that bodies found on the streets of Baghdad are up 41% from January?

From the graph above, it looks like civilian casualties are down 41% since last January, not up. How to reconcile? Different sources are used. The WA Post based their information on morgue data from the Health Ministry and Engram based his figures on the ICCC, which uses a different methodology.

A general comment. All I'm saying here is that June was better than May. This is only the first month of the surge strategy in full operational mode, so to cast any sort of judgment on the success or failure of the plan would be premature, to put it mildly. Things could easily go the other direction in the next two months.

"I'm afraid I'm not willing to say there is anything more than statistical noise on the charts presented."

The statistical fluctuation in a number is like sqrt(n), so for numbers like 1000, a difference over a hundred gets to be significant, so the first three plots don't have this problem. For quantities like 100 or 36, there is root-n is 10% or 17%, so the last two plots are noisy.

But all these numbers are in my view probably of little value because of _systematic_ uncertainties - maybe one expects fewer casualties in summer, maybe the reporting of incidents has been messed with, maybe civilian deaths aren't a good proxy for insurgent strength, etc.

Operation *Phantom* Thunder? Even the name admits it's a fake.

Quite a double standard on Iranian behavior. The Iranian seize 15 Brits in disputed territory, and it's a human rights outrage. American kill hundred or thousands of civilians by accident - nothing to see, move along now.

Of course, even if the "surge" manages to succeed (extremely unlikely) it's still an outrage that Bush did it. Any surge should have been done in Afghanistan, where the situation continues to deteriorate. Fighting in Iraq while Afghanistan remains unsettled directly attacks American security interests and deserts the struggle against Al Qaida (which is in Afghanistan and not in Iraq).

Statistics make little bit of relief,but still long way to go to get Zero-Death.But I wonder whether the casualties are gets down because of the population went down in the past couple of years..
Replacement Windows Suffolk

I would expect activity to quiet down during the extremely hot summer period.

More importantly, I think Baghdad is the wrong place to looking for a correlation between corpses and political progress. The Battle for Baghdad is probably winding down on the basis of its own dynamics, and the Shia won. Baghdad is for all intents and purposes a Shia city now.

It is a fundamental error to think that American activities are the primary variable for any set of statistics or indicators in Iraq. The Americans have been background players in the battle for Baghdad - it's their increasing irrelevance that is the central issue.

I still believe that, in a final bid to retain some meaningful degree of influence on the ground, they will bring back Allawi before the end of calendar year 2007. That truly will be the last roll of the dice.

"But all these numbers are in my view probably of little value because of _systematic_ uncertainties - maybe one expects fewer casualties in summer, maybe the reporting of incidents has been messed with, maybe civilian deaths aren't a good proxy for insurgent strength, etc."

I agree entirely. One of the most common errors in the use of statistics is to select a statistic which isn't really a good proxy for what you are trying to measure.

Cool charity deal that benefits wounded soldiers and their families.

What benefit has Charles Bird ever brought to this otherwise fine blog? Defense Tech has a shill like him who goes by "Christian". He seems to think that his role in life is to pump up anti-Iran hysteria. Consequently, I (at least) don't trust the site as much as I did when Noah Schachtman was running it. Don't let that happen here.

sglover : Don't let that happen here.

Your echo chamber is here, here, or here. Many others of course. If that is what you want in a blog….

What benefit has Charles Bird ever brought to this otherwise fine blog?

Quite a bit more than your bitchy commentary, glover.

And I agree with CB's last comment, even though I disagree with him on just about everything under the sun.

The statistical fluctuation in a number is like sqrt(n), so for numbers like 1000, a difference over a hundred gets to be significant, so the first three plots don't have this problem.

IANAS, but I think that this depends on the inherent noisiness of the data; you really can't have a 'statistically significant' difference with just two data points bc you can't see what the normal fluctuations are (eg if you roll a die and get a 4, then roll again and get a 1, it's less than 4-sqrt(4), but that doesn't suggest that the die is fixed).

Carleton, the sqrt(n) thing works for some things. If the distribution is gaussian, then it works real well.

And if you average pairs of samples to get your distribution, or average quartets of samples, etc -- the larger the number of samples you average together to get your revised samples, the closer it gets to gaussian given some assumptions like an infinite range and a finite variance.

So the sqrt(n) thing tends to get more true even for nongaussian distributions as n gets larger, provided they have infinite range in both directions and finite variance.

And even if they have finite range it still tends to work and it works better with larger n. Besides, finite range tends to give an even better result.

The only time it really fails is for distributions with infinite variance, and even then it works if you throw out the outliers. Of course, ignoring the outliers is kind of cheating.

So the sqrt(n) estimate works as a rule of thumb. It's likely to be reasonably close to the actual variance, assuming that there's a single distribution that your samples are drawn from. For reasonably large n, when different sample means differ by much less than sqrt(n) they probably won't be significantly different, and when they differ by much more than that, they probably will. In the former case you probably won't be able to tell they came from different distributions, and in the latter case you probably can tell they did. When it's iffy you can test more carefully or you can figure it's borderline and it might not be worth your time to get it clear.

It really makes sense to check for seasonal effects, but we only have 4 years of data so it's hard to be sure. And that data is corrupted by US choices. US and iraqi casualties go up when we choose to do big operations like fallujah and najaf. Whatever iraqi choices might be affecting the numbers will get swamped by US choices those months. So it would take some work to decide how reliably you can estimate seasonal effects.

Of course, statistical effects you can't explain are suspect, but then it's usually easy to come up with a JustSo story to explain whatever you find.

So -- rule of thumb. Rilkefan hedged further by using 100 instead of 10*sqrt(10). Yes, June is different from May. What that difference means is up for discussion.

The premise of this post is that crappy numbers are good if they are slightly less than last months crappy numbers (without taking into account the problems with whether or not the number counting is being done corectly). And allegedly this means that we are making progress and should keep at it.

Unfortunately, that logic is specious. Justification for pursuing the strategy should be based on an analysis that we are going to achieve something that merits the cost. So far, all that is being achieved is putting of the day of reckoning on this failure of Iraq policy.

Either that, or Charles will soon be writing about how lefties sold the nation down the river in Iraq just like they did with Viet Nam. Frankly, that seems to be the logic behind this post should the glorious surge be terminated before it has the next several years to realize an alleged victory.

Others have commented on the problems with the misuse of statistics in this post. I'm not a mathematician or a statistician, so I can't really comment from that perspective. What I can comment on is that even from a layman's perspective, the numbers at face value don't suggest that this is anything more than a blip.

To be more specific, compare June's numbers with October's of last year, or even with February and March. Notice anything striking? I was struck by the fact that June's numbers are either around the same or worse than those months, depending on the category. By your logic, does that mean the "surge" is making things worse?

Your numbers do not mean what you are trying to twist them into meaning.

"Your numbers do not mean what you are trying to twist them into meaning."

I am not sure that CB is tfying to twist them into anything. Several times during the post he points out that this is still early and things can go either way, and quotes Kilcullen to the same effect.

Soemtimes I think there is a kneejerk reaction to posts here, whether they be CB or hilzoy or von or publius.

I think, and perhaps I am being to generous, that CB is merely pointing out the current situation and that it might be indicative of a positive change.

As I mentioned in my first comment, my only quibble is with the term Democratic congress, since opposition is rapidly becoming bipartisan.

The premise of this post is that crappy numbers are good if they are slightly less than last months crappy numbers (without taking into account the problems with whether or not the number counting is being done corectly).

Sorry, dm, but that is not the premise of this post, dm, because I never said that the June numbers were "good". But thanks for misinterpreting. I also didn't say that we were making progress, so I'm not sure where you're getting all these weird revelations.

In my last post on the subject, I said that I would reserve judgment on the success or failure of the surge strategy until later this year, and I'm sticking by it.

BTW, the numbers are the best available, applied consistently from month to month. A question for you: why should it be OK to write a post on May but not on June?

By your logic, does that mean the "surge" is making things worse?

No, Catsy, because I didn't say here that the surge strategy was making things better in the first place. Last October, the strategy was different, so you would expect different numbers to result. Personally, I think the current strategy is better than the one under Rumsfeld, but for me it's too soon to judge how it's actually working.

Your numbers do not mean what you are trying to twist them into meaning.

Saying that June was a better month than May is "twisting"? Sheesh. By your own logic, why are you trying to twist the numbers between October and June?

So the "better" in "Better Month" does not equal "improved"?

Darn.

[...] But first, the numbers. Civilian casualties in Baghdad are down.

[...]

Civilian casualties elsewhere in Iraq are way down [....]

Etc. Charles, you keep applying the equivalent of body counts -- a lack of them, and using civilian casualties as a metric of success (despite the fact that the U.S. doesn't even keep track of casualties: what does that tell you? How do you explain this contradiction between your using this statistic as the most important one to pay attention to, while the U.S. government insists it's too unimportant to even measure?), when the issue is "has the Iraqi government gained in support and control of the country?"

You might as well be measuring jelly beans in a jar in the "Green Zone," for all that your numbers have to do with peace coming to Iraq. Sorry.

(Also, citing Some Guy On A Blog Who Disagrees With The Washington Post as a credible source of numbers, even if the numbers had anything to do with anything, doesn't fly, I'm afraid.)

"I hope the general can report sufficient progress to do so. Whether it'll happen, who knows."

Oh, for pete's sake. Yeah, sure, that'll happen, and it'll be accurate. Sure. And everyone will also have ponies!

Charles, what would it take for you to feel that you've reached a point where your credulity in believing We Still May Be Turning The Corner Soon has gone too far? This is an honest question, not an attempt to rag you: I'd like to know what you believe such a point would be.

"A general comment. All I'm saying here is that June was better than May."

How? How has the Iraqi governmental support and control improved?

dmbeaster: "Either that, or Charles will soon be writing about how lefties sold the nation down the river in Iraq just like they did with Viet Nam."

In fairness, Charles no longer, if he ever did, seems to hold such a view about Vietnam, although I leave it to him to clarify his views on that.

No, Catsy, because I didn't say here that the surge strategy was making things better in the first place. Last October, the strategy was different, so you would expect different numbers to result. Personally, I think the current strategy is better than the one under Rumsfeld, but for me it's too soon to judge how it's actually working.

Which I notice hasn't stopped you from making observations about how it's actually working, or from repeatedly pleading with us to give war a chance. As Model 62 noted, you describe June as a "better month" in the title of this post. The obvious implication here is that we are seeing signs of improvement, so we should give the "surge" more time. Yet when there are signs of the converse, we're supposed to give the "surge" more time because it's not in full swing yet. Your default position--pretty much since you started writing about Iraq, but particularly over the last year or so--has been "give it more time".

I mean, seriously. I get that you've evolved. Going from a (by your own description reliable Bush water-carrier to someone who takes a cautiously optimistic outlook on Iraq is progress, for which you are to be commended. The problem is that for the better part of the last few years you haven't substantially changed your tune. The only thing you've done is periodically pushed out the drop-dead date another Friedman or so. Back in November of 2005, your endgame date was mid-2006:

[American Forces Should Withdraw in Six Months] Why? Because their mission in Fallujah has been mostly accomplished.

When that turned out to be wrong, you next tried for the end of 2006:

The Petraeus plan looks to be one of the last and best tries. If we've made no discernible progress by this November, I may just put myself in the defeatist camp and call for a phased drawdown.

Most recently, you've conveniently punted the ball to next administration:

I think we should take that risk and make that final push until Bush's term expires. If we lose two years later, at least we can say we tried.

So at what point, Charles, do you stop moving the goalposts to the next six months? At what point are we expected to stop taking you seriously when you ask us to be patient and give Iraq more time? And at what point do we conclude that "more time" isn't going to lead to an acceptable solution in Iraq? At what point have we moved the goalposts defining victory so far that the definition becomes meaningless? At what point do we accept that deciding not to throw good money after bad is wisdom, not surrender?

...using civilian casualties as a metric of success...

Gary, you're the one using the word "success", not me. June is a better month statistically. The best we can say about civilian casualties is that they are a measure of the security situation, which is a reflection of terrorist-paramilitant-criminal activity and coalition attempts to keep it in check, and it is just one measure among many for evaluating how Iraq is doing.

You should know as well as anyone why the U.S. government doesn't track civilian casualties data, so I don't know why you are asking me. Is this a test?

Charles, what would it take for you to feel that you've reached a point where your credulity in believing We Still May Be Turning The Corner Soon has gone too far?

First off, your question is falsely premised. I haven't said that we've turned the corner and I specifically said "who knows if it'll happen". Had you read my previous post here, you would have learned that I haven't said "turned the corner" or the like for over three years. You might have also gleaned from it that I'm giving the surge strategy 'til the end of the year, that if we don't see clear and discernible progress by then, then I would declare myself a defeatist and call for a Plan B, which would most likely entail an orderly phased reduction.

Back in November of 2005, your endgame date was mid-2006

You're twisting, Catsy. That quote was a specific reference to Fallujah, not the entire country. In the very link you provided, I wrote that I would give the surge strategy 'til the end of the year, and have said so at least a half dozen times in at least a half dozen other threads.

When that turned out to be wrong, you next tried for the end of 2006.

No, that would be the end of 2007, not 2006, again from the very link you provided. Nice try.

Personally, I think the current strategy is better than the one under Rumsfeld, but for me it's too soon to judge how it's actually working.

What do you think the current strategy is, and how does it differ from the previous strategy?

I have some answers about that, but I suspect your answers might be different.

Oh well, I'll say a little bit about the old strategy. It changed at fairly regular intervals.

Insurgency? What insurgency?

Insurgency? What insurgency?

It's just a few Saddam dead-enders, they'll die or get discouraged.

They'll quit as soon as we catch Saddam.

Once we destroy the nerve-center in Fallujah the insurgency will be crushed. Also we need to kill this bad-guy shia, Sadr. His followers will stop acting up once he's gone.

OK, we need to rethink this. These guys are insurgents because they don't have anything better to do. Get a solid reconstruction going and they'll get jobs and make money and they won't have time for violence.

We have to get security first, then we can get reconstruction.

We don't have enough troops to provide security. Still, the iraqi people will rally around the democratic iraqi government and the insurgency will wither.

It didn't wither. So we'll move into insurgent-owned areas and drive them off, and then the iraqi army can keep them from coming back.

The iraqi army wasn't trained well enough to keep the insurgents out. So we'll train the iraqi army better. That's our mission, train the iraqi army. We don't need to take casualties ourselves fighting insurgents.

The insurgents are training faster than the iraqi army. So we need to train faster and better, and also fight the insurgents. Hey, we had a success in Tal Afar. This little minority group gets picked on by sunnis and shias both. They were *glad* to have us protect them. They cooperated as much as they could. They got upset when we moved 2/3 of the troops out to do something else. A wonderful success! Let's get the guy who won there to invent a strategy for the rest of the country.

I think too much is being read into this post. It's really not much more than a report on some numbers for June as compared to earlier months.

I think everyone agrees that the numbers are noisy and don't mean all that much. Still, a decline is better than an increase. I don't see Charles saying much more than that.

Another possibility is that we are going to start seeing fewer but deadlier attacks, leading to more variance in the month-to month numbers but not real decline in, say, three-month averages. This could happen if we get better at preventing smaller, less-carefully planned attacks.

By the way, it may seem minor, but adjusting for the number of days in the month would improve the accuracy a bit. This is less important than other factors mentioned, including variation, seasonal adjustment, etc.


Rilkefan and J Thomas,

I don't quite follow what is meant by

The statistical fluctuation in a number is like sqrt(n), so for numbers like 1000, a difference over a hundred gets to be significant, so the first three plots don't have this problem. For quantities like 100 or 36, there is root-n is 10% or 17%, so the last two plots are noisy.

What is "n" here? Are you saying that the variance of a measurement is, roughly, its square root? sorry to be dense but maybe you could clarify. Thanks.

What is "n" here? Are you saying that the variance of a measurement is, roughly, its square root? sorry to be dense but maybe you could clarify.

Bernard, yes, you got it -- for particular sets of assumptions.

Here's a sample set of assumptions that it would fit with.

First, let's suppose that there are around 50,000 combat troops in iraq who go out and get shot at. And let's assume that they each go on about 200 missions a year. So that's about 10 million person-missions a year. And say it's about 1000 US combat deaths a year. Then that means the chance of dying when you go out on a mission is about one in ten thousand. Now, how reliable is this estimate? It's more reliable when you have 64 observations than when you have only 16. 256 is better. 1024 better still. And the variance with this particular model is proportional to the square root of n. (The number of missions where the soldier doesn't get killed makes a difference too, but that's so large and varies so little compared to n that it doesn't much matter.)

Now, note that we've assumed a constant probability per mission, and we've estimated what that probability is. If there are some missions were people will probably die and some where they definitely will not, then you could get extra variance when the mix of missions changes. There are various ways for it o fail. But if the assumption is that the simple model is right, and n changes far more than expected from month to month, then that's pretty good evidence that the simple model is wrong though it doesn't say what model is right. That's what we have here.

We have various measures where the difference month to month is much larger than it would be if there was nothing changing. So we can reasonably say that something is changing.

If you want to make reasonable guesses about what's changing and perhaps show that the statistical evidence doesn't discredit them, that would take a lot more work.

What do you think the current strategy is, and how does it differ from the previous strategy?

Which previous strategy, J? The current one is at least based on well-developed doctrine and the U.S. commander is applying it. The Rumsfeld era was under-manned, ill-conceived and too slow in responding to evolving situations.

Charles, I mean the strategy we were expounding just before the surge.

We were going to get control in "oil spots" and then expand that control into bigger and bigger areas.

The theory was that we could move into an area and with careful searching, disarm it. And with careful use of checkpoints we could prevent them from re-arming. Then as the pacified areas could be turned over to iraqi police (and to some extent to iraqi army) we would have the troops available to expand into new areas. And the less violence, the more we could provide iraqi politicians with the room to get political solutions.


As I see the new strategy, it's mostly just the same. Except we're concentrating only on Baghdad instead of a lot of other secure areas. We're sending troops into anbar to stir things up, to keep the insurgents' rear areas from being too safe. And we're doing the same in the areas around baghdad, we're looking for car-bomb factories and generally keeping the insurgents from getting too settled.

And inside baghdad we're doing same-old same-old, with a few refinements. We think we can depend on iraqi troops more than we could before, so we're using them. We have two chains of command in parallel, which lets the iraqi CoC learn from the US CoC and see better how we do things, though they only have influence when we're ready to listen to them.

We're doing more foot patrols instead patrolling in armored vehicles. In theory that lets us interact with iraqis more. And it means a single IED won't take out as many people.

We're talking about doing more interaction with informers etc, talking to more iraqis. We've talked about that all along though.

I just don't see much that's new here. Even the idea of putting extra emphasis on Baghdad and letting the rest of the country go hang isn't that new -- this time last year we saw that Baghdad had deteriorated to the point that was necessary, and that securing the capital was a "must-win", that we were obviously losing when we were losing most of Baghdad.

What's new here compared to what we were doing this time last year?

I hope the general can report sufficient progress to do so. Whether it'll happen, who knows.

I know. It won't.

I'll be back in a couple of months to remind you that I said so.

The comments to this entry are closed.

Blog powered by Typepad