Here’s the title of a big new peer-reviewed paper in The Lancet:
Take your time and answer this question (you will be graded): TRUE or FALSE, scientists measured the effects of air pollution on mortality of a group of folks in Europe.
Come on. After seeing the words effects of air pollution on mortality, what else can you say but TRUE?
It is FALSE, of course. The three or four dozen researchers listed as authors never measured, not even once, the amount of air “pollution” any person was exposed to. Further, every single author knew that the title was false. And so did every editor.
So why was it allowed? What about the children!
No, it was our old friend The Epidemiologist Fallacy, a.k.a. the ecological fallacy. Nothing is better at generating papers—the currency in academia—than Old Reliable. Using it is vastly cheaper than relying on reality, which often lets scientists down (right, Gav?). I beg you will read the linked article to understand this ubiquitous menace and driver of scientism.
Not only wasn’t air “pollution” (dust, mostly) measured on individuals, but the proxies of air “pollution” weren’t even measured at the same time as mortality. And not only that, but, well, read the letter, which has it all.
The three of us submitted, fixed, and resubmitted a letter which explained the shortcomings of the Beelen et al., not asking it to be withdrawn—if journals withdrew epidemiologist fallacy papers, there would be oceans of blank pages—but to highlight the false claims made.
Alas, observation rarely trumps theory (right, Gav?). The Lancet decided not to publish and to let the paper stand, doubtless reasoning that since so many others used the epidemiologist fallacy, and got away with it, there was no reason Beelen shouldn’t, too. And anyway, it’s embarrassing to admit to error.
The epidemiologist’s fallacy – yet another example
Yours Truly
Pieternella S. Verhoeven
Associate professor at the Roosevelt Academy, Middelburg, the Netherlands
Head of the Institute of Undergraduate Research ‘Eleanor’, Middelburg, the NetherlandsJaap C. Hanekamp
Associate professor at the Roosevelt Academy, Middelburg, the Netherlands
Adjunct Associate professor University of Massachusetts, Environmental Health Sciences, Amherst, US
Chair of the Chemical Food Safety & Toxicity Working Group of the Global Harmonization InitiativeBeelen et al.’s paper carries a peculiar title considering that the authors never did what they claim: exposure to air pollution was never measured on any individual. It is only poorly guessed at, and not even guessed at over the right time.
The land-use regression models, which guessed the different kinds of pollution, are calculated using data from October 2008 through May 2011. Yet, the agglomerated studies ran from 1992 through 2007, with most from the 1990s. So even with correct pollution estimates, it would have had to operate backwards in time. Besides, it cannot be claimed that pollution from 2008-2011 accurately represents pollution in the 1990s because of weather dependency.
The “variance explained” by the land-use models is 57-89%, meaning the guesses are often wrong. Nevertheless, the guessed values at the participants’ residencies are presented as actual exposures, unreasonably leaving the participants ‘fixed’ in their residences within the study-timeframes.
Furthermore, the considerable error in the ‘exposure’ estimates is not encapsulated in the statistical analysis: the guesses are taken as fixed without accompanying plus-or-minuses. If that were done, the study’s results would have been rendered insignificant.
Additionally, the cities with the highest guessed pollution (Athens and Rome) had no and even a slight negative effect for PM2.5. Strangely, the one location (VHM&PP), which found a significant effect, was accorded the largest weight (ten times any other) in the meta-analysis.
Overall, the paper’s conclusions are obviated by making the exposure guesses error free, inflating one (of eight) slim results in favour of the proposed hypothesis.
You’ll like this as well then.
http://wattsupwiththat.com/2014/07/10/peer-review-ring-busted/
Briggs:
Excellent work!
Betty: ‘deep sigh’ “Who was that handsome man in the Borsalino fedora and Burberry trench coat?”
Mark: “Why Betty, that was Matt Briggs, AKA “The Thorn” [as in thorn-in-the-side of those he confronts]. He calls out and brings to justice all those who write and publish vacuous and untrue papers in the scientific journals.”
‘said seriously with an overtone of envy’ “Betty, he is truly a modern superhero.”
Betty: ‘swooning’ “Oh… he’s so wonderful! …Is he married?” ‘another deep sigh’
– Keep up the good work Mr Briggs!
I’m confused. Is this some kind of pro-pollution site?
JMJ
JMJ,
And what did you think of the epidemiologist fallacy article? And how it applied here?
JMJ: Interesting—one is pro-pollution unless one goes along with any useless study and fabricated data showing pollution is bad. Okay, we can create a study showing the use of any kind of perfume is twice as bad as car exhaust and finally we can rid ourselves of the scorge of perfumes. It’ll be a legitimate study with a wee-p value. Easy to do, since real people and real data are not actually required. I expect you to back me on this one or I’ll have to label you pro-pollution.
Here’s another charming study that covers both rediculous “causality” implications and global warming, at least according to the article:
http://americanlivewire.com/2014-07-10-shocking-warmer-climate-mean-kidney-stones/
Ah yes. If it can’t be measured directly, it can’t even be estimated, therefore it isn’t happening. Pardon me while I go stick my head in some bus exhaust for a few hours to celebrate.
Brandon,
Like JMJ, you misunderstand, probably because you didn’t read everything.
The pollution is NOT claimed to be an estimate, but said to be measured. The uncertainty in the estimate is NOT included in the analysis, which is therefore badly over-certain.
How did you like the part where they used estimates from one time period and mortality from another?
I thought the EPA had proved that you don’t have to measure any exposure to prove that some substance causes disease or death. The EPA claims that exposure to environmental tobacco smoke (ETS) causes thousands of lung cancers a year. The EPA didn’t measure any exposure to ETS. People were given a questionare that asked about their exposure. They were polled. According to the EPA you don’t need to measure any thing to determine causality, just take a poll.
BTW, there was a study that measured exposure to ETS in 1975, but it was ignored because it gave the wrong answer. http://dengulenegl.dk/blog/wp-content/uploads/2008/12/first2.jpg
Brandon: The EPA actually did the “breathe in bus/truck exhaust” experiment. You missed your opportunity to participate. (That study did have actual data, though there were rumours of no full disclosure to the participants. Maybe that’s why it’s easier to just make up data.)
Briggs, you are correct. I did not read everything.
My meta-frustration with your writings is that the bulk of them focus on things which have been done incorrectly. I cannot recall a recent post from you that pointed at something and said “this is how to do it properly”. Which may be a cognitive bias on my part. My initial reply to you today is certainly exactly that.
At the risk of asking for false balance, can you really not find an air pollution study, a climate paper, or the like which is interesting in the academic sense of the word, and which is done properly according to your statistical expertise? I for one would very much like to read a treatment by you comparing a “good” and “bad” paper in the same field. Good apples to bad apples if you will.
Regards, B
Sheri,
Yes, I missed the bus exhaust discussion, alas. My point is that common sense should dictate that breathing soot particles is unhealthy. Your point is that quantifying the hazard should be done properly. And I do agree.
Which obviously does happen. Food for thought: if all data were made up, then all models would predict them correctly.
Brandon,
If you can’t recall a post where things are done correctly, you haven’t been paying enough attention. Look on the Classic Posts page under Statistics, Probability, and Climate. Look especially at predictive methods and time series.
Brandon,
“My meta-frustration with your writings is that the bulk of them focus on things which have been done incorrectly.”
That’s because very very few researchers who’s work is dependent on statistics actually bother to consult with statisticians. Probably something on the order of 80+% of all papers that are dependent on statistical analysis did it wrong.
There is simply far more chaff than there is wheat.
Brandon: Good point on the made-up data. Maybe the problem is trying to hybridize enough read data with made-up stuff. Or maybe everyone didn’t get the memo on what the desired outcomes are? Who knows? Seems even fictional outcomes can be tricky. If we could just get researchers to learn proper statistical methods and research protocols, and accept that sometimes outcomes are not what one wanted, we might get back to good research papers.
Briggs,
This post reminds me of your analysis about morality in which you used a certain sum as a measure of morality.
Brandon’s suggestion is good. If you want to attract consulting work, I think you’d need to show how one can correct the problems, in addition to what others have done wrong.
This statement is incorrect! For this, I wouldnot accept the letter either.
In shooting down the statistical follies of members of the scientific community Dr. Briggs plays an essential role which, however, few of his colleagues are willing to play. Please keep at it, Dr. Briggs!
one is pro-pollution unless one goes along with any useless study and fabricated data showing pollution is bad.
The Calvinist nature of such an attitude is palpable. These papers are sermons; and the failure to shout Alleluia! is taken as evidence of one’s sinfulness. The religious nature of much secularism was pointed out to me once by a friend who was raised Dutch Reformed. He knew Calvinism when he saw it.
Briggs, take the studies as you please. Throw this one and a few thousand others out as well, you’d still have thousands of studies showing that breathing certain fine particulates in certain quantities over time is bad for you. The studies were liminal and agglomerated as such, but I don’t know of too many other ways we could get a look at the science there. But again, what kind of idiot thinks its a good idea to breath more pollutants?
JMJ
JMJ: Still not actually paying attention, I see. No one said pollution was good. No one implied that. This was not about pollution. It was about experimental methodology and statistics. Try rereading and see if you can catch that detail.
Actually, it’s too bad you too could not have volunteered for the EPA study where they pumped diesel fumes into a tightly sealed tube and let volunteers breathe it in to see what the effects were. (http://www.wncn.com/story/25155111/report-epa-fails-to-disclose-risks-in-human-tests) Maybe you could vent some of your anger on the EPA and their damaging tests. As you said “what kind of idiot thinks its a good idea to breath more pollutants?” I would answer that the EPA apparently thought it was okay to do so for the sake of science.
But again, what kind of idiot thinks its a good idea to breath more pollutants?
No sooner has it been pointed out that the bad science is secondary to the religious nature of affirming one’s belief, than Mr. McJones appears to provide a case example.
Briggs,
I’ve already read most of the classic posts in the categories you just now suggested. The time series posts are great, but I find myself yelling at my screen “for crying out loud, what about a scatter plot then?”
Is it me or have you expanded the list within the past several months? One article I had not read is this one:
https://www.wmbriggs.com/blog/?p=10601
Not just the tie but the three-piece grey pinstriper. All that’s needed is a fedora. Anywho, I’m with you on most of it until the last graf:
Now to review, Monckton’s statement was:
I don’t see why his conclusion must follow. What I do remember from stats class is this: when hypothesis testing, if the statistical test returns a result which is not significant, all that can be said is that the data do not support the hypothesis. And nothing else. Most definitely NOT that the hypothesis is wrong.
Where have I gone wrong here?
Sheri,
Ok, you’re begging the question that there is a conspiracy. Who knows?
The most damning evidence out of E. Anglia was “Mike’s Nature trick to hide the decline.” You’d think that years of e-mails from “researchers” tied to a Gore and Soros conspiracy would contain some mention of desired outcomes. No?
Keep in mind that this is an international conspiracy you’re talking about here, so what about Eric Snowden? Do you really think The Guardian would have sat on a scoop like that?
I have no idea how many FOIA requests have been filed over the years. They generate some nose-wrinkling results amplified by insinuations and to-and-fro mudslinging, but no one single smoking gun which shows that the data are clearly being fudged in a systematic organized process.
The code and data for many many of the climate models used are publicly available. And they show results the contrarian community calls “wrong”. What kind of ineptitude botches a conspiracy to such magnitude yet simultaneously manages to keep it all but under wraps?
MattS,
That means that 20% of the paper out there do it correctly. Why doesn’t Briggs write about them?
This is just ridiculous. The Heartland crowd just wants regulation-free polluting.
JMJ
It does not follow immediately. Example: an analysis of tablet potency versus tablet weight revealed a lack of correlation. But this flew in the face of reason. A larger tablet must contain more of the active ingredient, and so score a higher potency. The “lurking variable” was raw powder batch. The powder was prepared in batches by adding the right amount of A to the right amount of B. The A (for active ingredient) was assayed at incoming. If the incoming batch scored higher than the target 100, then the blending room added less of it to the batch make-up. If it scored lower, they added more “to make up for the weakness.”
However, most of the variation in the raw assays was due to measurement system variation — one of those pesky variations that no one wants to talk about — so even if all the incoming raws were identical in potency, they would still jigger the recipe to “counteract” the (dare I say it?) random variation in the measured amount.
Consequently, different powder batches had different potency due to adding more or less of the active in the mistaken belief they were correcting for a variation in the raws. In the correlation analysis aforesaid, the correlation was masked because the data covered two powder batches and the batch to batch variation resulted in the two “hotdogs” blending into a “hamburger.”
Thus, lack of correlation need not mean lack of causation. It may only mean sloppiness.
An apparent correlation between Y and X may be because of a lurking Z, which is causally connected to both X and Y. It may also be because Y is a cause of X. For example, higher surface sea temperatures result in less CO2 being held in solution by the oceans and the excess is released into the atmosphere.
++++++++++++
a) Pet peeve: That is not “begging the question.” The death of logic and reason at the hands of zeal has led to muddying the terms of logical art.
b) “Conspiracy” is not an issue. The appearance of conspiracy can be achieved though evolution by natural selection by means of funding control, editorial control, groupthink, and the like. Go along to get along is not a conspiracy. We saw groupthink quite often among managers in business and industry. Cholera-victims need not conspire to spread the disease; they need only drink from the same contaminated well.
Example: successive replications of Millikan’s oil-drop experiment resulted in higher and higher estimates of the electron charge. This was because Millikan had underestimated the charge in his original experiment by discarding some of his data that didn’t look right. Later experimenters got higher values but because they “knew” Millikan had done the crucial experiment, they fudged their results downward — just a bit. It was not a “conspiracy” to “hide the incline” but rather groupthink plus belief in the consensus. Eventually, facts won, as they have a way of doing.
“That means that 20% of the paper out there do it correctly. Why doesn’t Briggs write about them?”
Because his intent is to shame the ones doing it wrong until they see the error of their ways.
Brandon,
Absence of correlation necessarily implies absence of causation. Game over, logically speaking.
I don’t see why his conclusion must follow.
Because causation necessarily implies correlation. The absence of correlation means there is no link between the variables. Did you, by any chance, read the book I linked?
Brandon, you have a fantastic idea. Start a blog reporting about the 20% and watch the traffic drain for this site. Otherwise …..
Brandon: Check out http://stevengoddard.wordpress.com/2013/07/28/understanding-nasanoaa-temperature-adjustments/
There are a lot of other sites referring to this problem, also. The way you can ascertain deliberate (whether it be conscious or not) manipulation is due to the direction of the adjustments. When the majority of adjustments help your cause, there is reason to further examine the data. Which is what was done. It also appears NASA has been interpolating data for temperature stations that have been abandoned. Now, it is acceptable to interpolate data IF you have very strong evidence that your algorithm matches reality. You have to have tested the algorithm thoroughly. There does not seem to be any such testing.
YOS addressed the causality/correlation part. I was very surprised when a commenter on a blog I was reading said you don’t have to have correlation to have causality. It turned out to be an argument similar to those YOS mentioned in his comment–Z was lurking and not known. So there was correlation, but not between the variables the researcher thought. There has to be correlation for causality.
As for studies done well, the idea is to learn what is wrong with studies and then apply that to studies you read. There are cases where it’s hard to tell and cases where much more research of statistical methods will be needed. I see this all the time by reading both skeptic and warmist blogs–graph wars, comments about how wrong the other sides statistics are etc. It’s a learning process, and as Deebee pointed out, people tend to find error identification more interesting than “the right way” demonstrated. For some, that can be frustrating.
YOS;
Or in the scope of the climate system, many many lurking Zs.
And that is demonstrably the case. It need not be that X, Y and lurking Zs cannot all effect each other, as any basic understanding of the physics involved suggests.
So what’s your actual argument here?
Facts need to be gathered. We have rather sparse coverage of the largest heat sink in the entire system.
DAV;
Have you by chance figured out that correlations are numerical and can take more than just “yes” or “no” values?
DEBEE;
http://realclimate.org/
MattS;
So we can just go ahead and trust the 20% of the papers already doing it correctly then. Thanks for confirming that for us.
Sheri;
I know it has been done because I have done it myself. I downloaded every scrap of raw and adjusted station data I could get my paws on and did my own comparisons. The differences were negligible. My results looked quite similar to this:
http://www.skepticalscience.com/print.php?r=255
The reason for the spatial interpolations for sites that have moved is for the GCMs because they rely on geographically gridded information. There is no other way to do it.
It’s good to question and investigate, but in the end you need to read what the actual researchers themselves say they are doing and why:
http://data.giss.nasa.gov/gistemp/
If you cannot bring yourself to trust the people doing the actual work, then you may as well make up whatever the heck you want to believe. The only other option is to get a PhD in a relevant discipline and grab a thermometer.
Have you by chance figured out that correlations are numerical and can take more than just “yes†or “no†values?
Ya think?
Regardless, when it was said: Absence of correlation necessarily implies absence of causation. Game over, logically speaking. regarding which you said: I don’t see why his conclusion must follow. you are indicating total misunderstanding of the relationship between causation and correlation. There may be degrees of correlation but when it is absent there can be no causation.
Agreed establishing absence may be difficult especially when sampling alone can cause apparent correlation between two clearly unrelated variables.
Considering that the model outputs are far more correlated to rising CO2 than the observed temperatures have been with increasing divergence clearly shows the dependencies assumed by the models are overstated therefore the models are unreliable. So, yeah, for the current models and all the “suggested” courses of action based on them, Game Over.
Testing for moderation consumption of comment. Please ignore.
Okay, for now I’m out of here, there is hopefully a response to Brandon stuck in moderation and not eaten by cyberspace. Looks like I need to do a statistical analysis of my posts and see if I can locate the “moderation catch words”!
Trying again:
Brandon:
Again, you have come to a different conclusion than I have and are assuming I do not read the same things you do. I did not find the differences “negligible” and the shenanigans by NASA look very, very bad. Yes, it’s gridded. Yes, you want data in the grid. Solution: Make sure you have the measurements being taken there. Or you end up fabricating a lot of data—which is the problem. There’s far too much fabrication and too little real data.
Real Climate is not really a place to learn about the methodology and statistical accuracy of studies. There’s little discussion of the short-comings and a lot of those 80% studies over there.
I am getting very tired of being told to “read what the researchers themselves say”. I do, I did, I took a complex class to understand the calculus and physics basis of the research. You’re starting to sound like a “typical warmist”, which is not good. Discussions end when you start accusing everyone who disagrees with “not studying the data”. I can tell how the gridding for temperature is set up, why they used the grid they did, etc, but I will also tell you that interpolation is a crappy way to get data and this needs to be addresses.
“Trust the people who do the work” is a very naive statement. You and I both know there is a huge amount of uncertainty in all of this and very good researchers are not necessarily going to agree. So enough with the insults and “you just don’t trust” mantra. Some of us do read and study and learn and still disagree. You’re allowed to complain if I suddenly say the gravitational constant has changed because researcher BOB said so and he won’t share why. Otherwise, in dealing with statistics and probabilities, enough with this “you didn’t read” statements.
Sheri,
Really? I have a different impression about Real Climate. I am not talking about the comments there. Many posts there discuss papers published by both sides. Have you tried to read those papers? Which one? I would love to discuss the statistical methods in any of those papers with you. No imaginations of what/how others think. No personal attacks, just statistical methods of your choice!
—
I don’t know why people take Drol-l Monckton seriously. If he bills himself as a comedian, it is a different story. What does “correlation” mean to him? Is it a statistical term? If it is a statistical term, he is obviously wrong. For example, Pearson correlation coefficient can be 0 when there is a certain perfect quadratic relationship between two variables. Whether there is a causal relationship will depend on the context. If it is not meant to be a statistical term, the correlation is what researchers are looking for. If one can help researchers to do a better job, he/she should email suggestions to those researchers, instead of, making strange noise about their research. Researchers are not as naive as deniers want them to be. But if you want to engage in fight with blogger activists, have fun (though I think it is a waste of time)!
I’ll see what I can find on Real Science to discuss.
I suppose people take Monckton seriously for the same reason they take Al Gore seriously.
While you consider it a waste of time to engage in “fights” with blogger activists, I find it very useful for understanding the viewpoints of all sides.
Should be Real Climate. Best I wait until tomorrow to find a topic as my attention span and “paying attention factor” are both far below par today. I’ll check back.
Hi. JH
You’ve illustrated one of the problems with Pesrson correlation where you are told no correlation when there is in fact. I personally prefer mutual information which measures the amount of information shared between the vvariables. Normalized it’s called redundancy.
I believe Monckton is referring to the obvious discrepancy between the model predictions and the last 17 years of temperature record. One can visually confirm the lack of correlation — no math required.
Sheri,
Responses to your response to JH:
That just warmed the cockles of my lil’ ol’ heart.
Amen.
Responses to your responses to me:
Ideally you would want the measurements to have the same spatial density across the entire region being measured.
http://www.ems.psu.edu/~nese/world-st.gif
Do you really such a thing is feasible? In polar regions? The smack dab middle of the Sahara? Deep in the bush of the Amazon rain forest? In that great big hole of the south east Pacific Ocean where there are no islands?
The dark arts of interpolation aren’t typically used because of a need to “fabricate” data. We need to interpolate because we must be able to use the data we have. If every researcher sat around waiting for the data they wanted, zero research would ever get published. In any field of study.
Reality is not partisan, Sheri. I am an empiricist. Sound theory matched with the best available evidence is the lens through which I view the world. None of my formal training is in climatology. I’ve got basic physics, calc and stats. I have to rely on domain expertise to help me understand complex problems which lie outside my own training — which when compared to the entire world of knowledge is pretty much nil.
I don’t pooh pooh the coursework you’ve taken. I think it’s fantastic that you have done so. When I say, “read what the researchers say” I am not being at all critical of what it is that you have taken the time to learn about. I’m being critical of an attitude which permeates the climate sceptic community, which is: climatologists are a bunch of ignorant dolts missing this one obvious little thing blah blah blah blah.
Monckton is the very embodiment of just that sort of hubris. Let’s just read this together one more time:
Come off it already! There is something basically wrong with the man if he actually thinks that a PhD in any physical sciences field doesn’t understand something that everyone in a first year undergad stats class learns. It’s such a preposterously transparent piece of disingenuous partisan crapola that it insults reason itself, never mind the community of researchers he’s maligning.
And you of all people should know by now that I don’t trust blindly.
Let’s talk about this insults thing. Read this blog from my perspective sometime. When I cork off some snark, or vent some frustration — can you honestly tell me that it’s wholly undeserved? Do I really seem the type to hand out gratuitous insults when I’m a guest in someone else’s house?
Do you know the reason that I speak plainly to you is for the very reason that you’re so straight with me?
DAV,
Which is a patently false representation of my understanding. Which is exactly how I know that you are full of crap.
Correlation is NOT absent. See again about this not being a yes/no proposition. When you’re ready to be an honest sceptic, you’ll stop saying such silly things.
JH,
On the whole that site does about the best job of any blog keeping an eye on the science and staying out of the politically (and ego) motivated mudslinging. Not perfectly of course, such is just not possible on this topic. The comments are much better than average as well. I find the mods to be particularly even-handed in their dealings with sceptics and denier-trolls.
My favourite sceptic blog has to be Judith Curry’s. Possibly my favourite post ever is this one:
http://judithcurry.com/2014/05/26/the-heart-of-the-climate-dynamics-debate/
Which was a response to:
I had never seen both positions so concisely stated and in the same place. And then Judith,
That’s how to do a discussion. It’s the essence of what it means to be sceptical seeker of truth. I almost wept reading it.
Which is a patently false representation of my understanding. Which is exactly how I know that you are full of crap.
Oh, I dunno,
When it was said: Absence of correlation necessarily implies absence of causation. Game over, logically speaking and then you said: I don’t see why his conclusion must follow you are saying the former is logically false (which is what “does not follow” means) — demonstrating lack of understanding.
Have you by chance figured out that correlations are numerical and can take more than just “yes†or “no†values?
The conclusion about the absence of correlation wrt to causation is s correct regardless of any particular case.
Correlation is NOT absent.
By all means, do show it is not.
http://www.drroyspencer.com/wp-content/uploads/CMIP5-90-models-global-Tsfc-vs-obs-thru-2013.png
So, looking at this with the model slopes near 30 degrees with the temperature record essentially flat for a minimum of the last ten years then saying there IS correlation you are continuing to show lack of understanding. Assuming the correlation is non-zero (which it nearly always is in any study), it must be quite small to the point where it is unimportant and it’s quite clear the models have it wrong and are thus unreliable. You are grasping at straws. Is that why you feel the need for “crappy” statements? Is this what you meant by That’s how to do a discussion
Editing using HTML tags is a pain.
Sheri, we are looking for different things, aren’t we? Have fun, as I said.
Hi (from Beijing), DAV, I have no idea what Monckton was talking about. Does he know what he was talking about?
Hi JH,
Well can’t say for sure but he seems to.
DAV,
I don’t see why his conclusion must follow. As in why it must necessarily falisify the theory. Hidden variables anyone? Quasi-periodic oscillators which can’t be reliably predicted — like ENSO perhaps? Crikey. Oh, I forgot, nobody pays attention to anything but CO2 in your little lab bench world.
So we’ve gone from discussing CO2/Temperature correlation in the instrumental record to Spencer’s graph again. Way to move the goalposts. So you want to discuss model performance to temperature records? How about something that goes back to 1880 instead of only 1980?
http://static.berkeleyearth.org/graphics/figure9.pdf
Oh gosh. what a horrible correlation. Nobody knows what the heck they’re doing.
Brandon: I fully believe that using insufficient, improperly obtained data is as bad a science as one can get. The worst. If we don’t have the data, we don’t have conclusion except an imaginary one in our little minds. Science requires data. This is one of my big problems with global warming–there isn’t enough data and where it doesn’t exist, they interpolate without any evidence that the interpolation is accurate. So to answer your question, if it’s not feasible to measure the temperature globally, then it’s not possible to know what the average global temperature is nor is it possible to know if that average is changing. It simply isn’t.
No, we do NOT have to use the data we have. If the data is incomplete, we shouldn’t be using it at all. That’s the problem with this.
How is your attitude towards skeptics different from your perceived attitudes about what skeptics see in climatologists?
Enough with Monckton–I have clearly stated everywhere that I consider Monckton to be skeptic’s version of Al Gore. I’m not going to discuss Monckton unless we throw in Gore and have a free-for-all.
I can tell you that when you “snark off” at people who are not committing the fallacies and don’t hold typical viewpoints, you are alienating people and cutting off any source of learning that you say you are interested in gaining. Yes, we are straightforward with each other–which since we both are fairly tough works. I don’t consider somethings being “straightforward”–like telling me to read up and study. That I consider insulting, since you know I have read at least as much as you. Are you telling yourself to read up? Not being snarky here, just asking? You also should know by now that if convincing evidence arose, I’d change my mind and that I keep reading and studying, so the comment to me looks like snark, pure and simple.
I do find it disturbing that you find Real Climate to be “one of the best blogs”. Three of the seven or eight contributors are the top political activists in the field and I have a tremendously hard time sorting through politics, activism and actual science. I couldn’t locate a study (actually, studies that aren’t just abstracts are hard to come by) showing “the other side” with is not surprising since both Schmidt and Mann clearly state there is no other side. Judith Curry is good and seems to try and avoid politics. (As for the argument from authority……probably best not to go there at the moment)
I will note concerning your response to DAV–backcasting generally works because models are “trained” on it. If the model can’t even follow it’s training……Predictions are where every model is very, very weak. This is stated by climatologists as being the inherent weakness of the models, which is interesting since without prediction, they’re pretty much worthless.
So we’ve gone from discussing CO2/Temperature correlation in the instrumental record to Spencer’s graph again.
Hello??? The graph is WHAT Monckton was talking about! Or at least is clearly shown by it.
How about something that goes back to 1880 instead of only 1980?
Uhh, the divergence between the model forecasts (which seem to be tracking CO2) and the temperatures begins ca. 1988-2000 IIRC.
Oh gosh. what a horrible correlation. Nobody knows what the heck they’re doing.
Strange reflected image.
Ask yourself when the models were most recently tuned. They are made to backcast, i.e, agree with the past. The number one paper you linked umpteen days ago shows you how they do it (more or less). And because they agree (also more or less) with the training data it is gleefully trumpeted they must be right.
If you haven’t already seen it,
Roy Spencer’s ICCC9 speech might be worth watching. I’m only halfway through it now. I had a lot of trouble getting to it. The only linux machine that seems to be able to display it has no sound interface so I’m watching it on an old Vista machine. Haven’t used IE in a long time and am surprised at the number of ads that I never see on my linux machines.
For whatever reason this site and WUWT are really slow compared to say CNN. Maybe Verizon has implemented QOS? Or maybe I’ve picked up a couple of hitchhikers on my Broadband Highway (like what happened some years ago).
Anyway watch the video.
Sheri,
Remember that all models are wrong; the practical question is how wrong do they have to be to not be useful. ~George E. P. Box
Yes, that would be a shameless appeal to a Bayesian authority. Are you asserting that all data are improperly obtained? What resolution of coverage would satisfy you? One thermometer per square kilometer of surface? Per cubic kilometer of ocean and atmosphere?
Sheri: stuck in moderation, and looks like I screwed up a blockquote tag.
DAV,
I wanted to test AGW theory that rising GHG concentrations have been causing rising temperatures. A simple model at annual resolution seemed doable. So I went out and grabbed a bunch of data going back to 1880:
http://data.giss.nasa.gov/gistemp/graphs_v3/Fig.A2.txt
http://data.giss.nasa.gov/modelforce/Fe.1880-2011.txt
Comparing GISS temp to GISS forcings is apples to apples. But any other global temperature record will do.
You can get CO2 concentrations from Mauna Loa from 1959-2013 at this link:
ftp://aftp.cmdl.noaa.gov/products/trends/co2/co2_annmean_mlo.txt
To go back to 1880 for CO2 from the Law Dome ice core data:
http://cdiac.ornl.gov/ftp/trends/co2/lawdome.combined.dat
The 20 year smoothed ppm series would be my choice.
Forcings without taking internal variability into account doesn’t do much good:
http://www.cgd.ucar.edu/cas/catalog/climind/SOI.signal.ascii
SOI is generally regarded as the biggie, but there are many others. It’s also fiendishly difficult to predict. So my model would theoretically do well over training data, but fail miserably if it couldn’t adequately predict those kinds of massively significant and difficult to forecast oscillators.
But that doesn’t stop me from testing the hypothesis that rising GHG levels account for at least some of the rising trend in surface temps since 1880. But I don’t want to cheat and jimmy GHG concentrations to fit my temperature curve. So I turned to this tool:
http://www.nutonian.com/products/eureqa/
It’s free for 30 days and allowed me to run three experiments:
1) Generate a model using only insolation, aerosols and SOI.
2) Generate a model using only CO2 concentrations. (Not the GISS calculated GHG radiative forcing, but the actual CO2 concentration.)
3) Generate a model using all parameters except the GISS calculated GHG radiative forcing.
Note that only the second experiment is the only one where I explicitly asked the package to look at CO2 concentrations. I otherwise gave it no other hints.
Brandon:
No, I am not asserting all data is improperly obtained. Currently, the best resolution I have read is 5 km squares. This does not allow for modeling of clouds, etc. but rather parameterization. If we can’t model clouds, that would seem to be too little resolution since clouds are a major factor in climate. I’m not sure what resolution would satisfy me–temperature is so widely varying that trying to obtain enough data to actually “know” the average global temperature may be a fool’s errand. I’ll have to think that over.
Complete data would be stations that have data throughout the period we are analyzing. If a station is lost, then it’s lost. No interpolating to fill in. Interpolating temperature values is neigh unto impossible as temperature varies over very, very small distances. I would have to see evidence that the interpolations are accurate–as in an algorithm that can accurately produce the temperatures via interpolation at stations with known values.
It would take more than just a line that shows the temperature rising. The analysis would have to be able to handle variables like El Nino and La Nina, volcanoes, ocean currents, clouds, etc at a resolution not requiring interpolation. Yes, that means copious amounts of data. The model would have to be able to predict reliably, if we are using a model. Paleo data would have to be shown to be accurate at a level that allows us to calculate mean temperature to tenths of degrees.
I don’t read scienceblogs or rabett much nor tamino. Same for Watts. I actually read more papers than blogs, though I do like scienceofdoom because they don’t get political very often and cover the math and theory of climate science. I rarely read Real Climate because of the mess of politics, etc. I don’t like sorting through to find whatever science might be hiding there.
Agreed that JH providing an example would be helpful, since I really don’t know what was being referenced.
Wonder if we were bad and that’s why we keep getting pitched into moderation? You’d think after we were killed off in the play we would be pretty much harmless now! 🙂
Brandon: Your examples on contentiousness in climate science were interesting. I would say that while gravity isn’t a useful theory because Newton was nice, had Newton been as agressive and threatening as some climate scientists are, his theory would have died a slow death until someone with more people skills came along. Science is determined by “niceness”, but people accepting science certainly is. At least there was no mention of “science is a blood sport” which I’ve read as a quote.
Brandon Gates 14 July 2014 at 8:34 pm
It’s not at all clear what you were trying to say.
You grabbed some data and found a tool (which looks like a genetic algorithm engine) that allows/allowed you to do three things and then ????
Whatever it was and given how many words you used to say only the above sentence you probably should write it up and publish it somewhere (maybe Briggs will take it as a guest post) as it looks like it would take up far more than could/should be handled in a comment. And when you do publish it, try not to ramble and don’t stop in the middle of the story. You say you read a lot of papers. Maybe you could use one of them as a guide for how to go about making your presentation.
Brandon:
Like DAV, I wondered where you were going with this. You’re showing correlation, but not causality. There are a number of assumptions in the example–The CO2 concentration at the Mauna Loa volcano represents the true average concentration of CO2 on earth, and that ice cores are somehow an accurate proxy which can then be tacked onto the instrumental record because of there is no differences in resolution. Those two assumptions are quite large.
Now, since you were saying it accounted for only part of the increase, there is a chance the data could be useful, if you can show the causality. If you use only the CO2, that does not prove causality because of the very real probability of important missing variables in the model. If it’s only CO2, then we’re wasting our time with supercomputers, and I know you don’t mean that.
DAV/Sheri,
It’s an experimental protocol. I know, I know, you hate my homework assignments.
Sheri,
And as I understand it we can’t model clouds to anyone’s satisfaction. Not only is cloud albedo important, precipitation obviously depends on them. Not getting that right, or even close to right, would be a fail in and of itself.
Ocean circulation is the other big challenge I’m aware of. Obviously we’ve not gotten it yet on an annual or even decadal resolultion, otherwise the models wouldn’t be so far off over the past decade.
It could be a fool’s errand, but that depends on what the errand is. If you want accurate annual predictions (all outputs, not just temps) +/- 1% down to a 5 km grid, forget about it. Decadal predictions on a 50 km grid for planning purposes and risk analysis are more doable.
Certainly we know enough to say that CO2 is raising earth’s equilibrium temperature. We don’t need hyper-skilful GCM backcasts to know that. We know that we would be wise to step on the brake, but many of our friends on the left are being rather obtuse about how to do it — NO NUKES, etc. So I’m not entirely surprised that so many on the right balk at them.
There’s a reasonable debate to have on what to do, but that can only happen if there is reasonable agreement on the science itself. Which there isn’t … at least not officially.
If anything other than a big meteor from space wipes us all out, I’m fairly certain that politicians will be the ones to do it … they already have quite the historical track record.
Brandon: If I get time, I’ll try the “homework”.
Yes, we “know” CO2 raises temperature and we are putting it in the air. That’s really all we know, since we don’t have an adequate understanding of the interactions of all parts of climate. Think of it as we know Drug A causes a certain set of side effects. We give Patient A the drug and a side effect connected to Drug A does appear. But wait–Patient A also is taking four other drugs and we have a very limited understanding of the interactions. Can we conclude Drug A is the cause of the side effect?
I certainly agree that if the fear of nuclear and the political climate weren’t such a problem, we probably would and could reduce CO2 output, whether or not it’s warming the planet.
It’s an experimental protocol. I know, I know, you hate my homework assignments.
Doing it yourself ever occur to you? Go do it; write it up then we have something to talk about.