Bishop Hill A $100,000 climate prize
Nov 18, 2015 Climate: Statistics Climatologists often claim that they are able to detect the global warming signal in the temperature records. If they are right then they are going to be having a very happy Christmas indeed, because Doug Keenan is offering them the chance to win a very large cash prize at his expense. Here are the details.
There have been many claims of observational evidence for global-warming alarmism. I have argued that all such claims rely on invalid statistical analyses. Some people, though, have asserted that the analyses are valid. Those people assert, in particular, that they can determine, via statistical analysis, whether global temperatures are increasing more that would be reasonably expected by random natural variation. Those people do not present any counter to my argument, but they make their assertions anyway.
In response to that, I am sponsoring a contest: the prize is $100 000. In essence, the prize will be awared to anyone who can demonstrate, via statistical analysis, that the increase in global temperatures is probably not due to random natural variation.
The file Series1000.txt contains 1000 time series. Each series has length 135 (about the same as that of the most commonly studied series of global temperatures). The series were generated via trendless statistical models fit for global temperatures. Some series then had a trend added to them. Each trend averaged 1°C/century—which is greater than the trend claimed for global temperatures. Some trends were positive; others were negative.
A prize of $100 000 (one hundred thousand U.S. dollars) will be awarded to the first person, or group of people, who correctly identifies at least 900 series: i.e. which series were generated by a trendless process and which were generated by a trending process.
Each entry in the contest must be accompanied by a payment of $10; this is being done to inhibit non-serious entries. The contest closes at the end of 30 November 2016.
The file Answers1000.txt identifies which series were generated by a trendless process and which by a trending process. The file is encrypted. The encryption key and method will be made available when someone submits a prize-winning answer or, if no prize-winning answers are submitted, when the contest closes.
More here.
Reader Comments (129)
Now, let's see just how creative these warmists can really be! Meantime, I'll enjoy the Interglacial, although having said that, I have noted in the last month or so & particulalry in today's Wet Office forecast, they are softening us up for cold stuff! Not one mention of "how unusual this event is", especially as snow is a thing of the past our children & grandchildren will never experience! Ooops, there goes another rib!
Okay, this does grate a little, but I think Doug Keenan may owe Richard Tol $100000. Richard's 1993 paper - Greenhouse Statistics-Time Series Analysis concludes with
Richard's 19971 paper - A Bayesian statistical analysis of the enhanced greenhouse effect concludes with:
I would argue that that is a statistical analysis showing that the increase in global tempertures is probably not due natural variations. Maybe Richard will share some of the prize money with me?
To be more serious, though, this would only make sense if the trendless statistical models were actually representative of natural variability. If they are simply trendless statistical models that - in some cases - happen to match the observed surface temperature time series, then they almost certainly are not. As Richard Tol's 1997 paper says
So, at best, this competition is trying to establish if one can distinguish between time series' with a trend, and without a trend. That tells you absolutely nothing about whether or not the increase in global surface temperature could be due to natural variation, or not.
I would have put $100,000 on ThenThere's to be the first to attempt a debunk.
As usual, little physics in his post.
The point is we don't know the nature of the natural process. The whole point is being able to discern between an unknown, but zero trend, process and a deterministic (i.e. a mathematical function of monotonically increasing CO2) trend.
Go to it Physics. Go get loaded!
I meant to add.
Well done Doug. An excellent and definitive test.
Even if you gave an infinite number of climate scientists an infinite number of computers with random number generators, a 12 month timescale seems bit tight. Some of them have been doing this, since before the hiatus started. Some of them for even longer, since before unreliable dendrochronology was invented over a cup of coffee.
What's more interesting is that the encrypted answers file is 500 bytes long. Perhaps it's a gzip of a comma delimited list of 0 and 1 answers for 'trend' and 'no trend, which was encrypted and base64 encoded.
If one abstracts both signal and noise, one still has to account for the slope left behind where the error bars used to be:
http://vvattsupwiththat.blogspot.com/2015/03/better-than-best.html
As this is true oof all the historical time series in queston, please remit $100,000
Totally confused, me. Demonstrating that a time series <I>has</I> a trend surely tells you nothing about the <I>cause</I> of the trend does it?
Mind you, I don't understand how a single time series can have an 'average' trend so what do I know?
I'm sorry, but is this supposed to be serious? In what world is this true:
Global temperature series are not 135 data points long. They're much longer. They would only be as short as Doug Keenan is claiming if one were restricted to annual temperatures, a completely unreasonable restriction nobody genuinely interested in studying global warming would apply.
Yes. ATTP could be making himself 100 Grand, at the expense of a person whose views he is strongly opposed to. But he chooses the low road instead.
The great thing is, Doug Keenan knows that people like ATTP might spend the time tying to find him out, which means they have less effort to spend on sabotaging Western economies built on the industrial revolution.
Do not underestimate the power of hypocritical conceit in academia. Especially in science, where it is common to think of oneself as being over-intelligent and under-paid.
Don't get me wrong, most are real humans who really would like to improve the lot of humanity through their best efforts, other things being equal.
I should also caution that the people who set out to prove Doug Keenan wrong may end up teaching themselves something they didn't want to learn.
The prize is going to be awarded for identifying that a subset of artificially created time series extracted from the total have a trend (+ or -). How is this going to determine that any discovered mathematical trend in a time series in the real world has any particular cause, or even multiple causes? Have I mis-understood something here (my stats education is 60 years past!)?
Tony.
Count me among those who don't necessarily equate Squiggology with reality.
Andrew
michael hart:
I think you have to be incredibly naive to believe this "test" is one which could actually be solved.
I suspect all they'll learn is what anyone who understands the basics of analyzing time series already knows - given enough model parameters, you can create a noise series indistinguishable from series with low levels of information.
If you treat the entire temperature record as only having 135 data points, as Doug Keenan has done here, it's relatively easy to create patterns of noise indistinguishable from it. I don't know what that's supposed to tell us though since the real temperature record is nothing like what Keenan's test series are like.
It looks as if ATTP has already made much the same point as mine. That does not necessarily mean that it is a silly one.
Tony.
You know, while my previous comment is dismissive of this test as having any practical meaning, I have to wonder if Doug Keenan would actually be willing to pony up the $100,000. If so, it might be interesting to see if one could solve the test just to try to get the money. I don't think the test itself has any real value, but $100,000 obviously does.
I just can't help but be skeptical he'd actually pay the money. That's a lot of money, and there's no mention that it's been deposited in an escrow account or anything. Do we have any assurance it'd actually be paid?
Even though the intention of this may be to get people to actually work out the answer, rather than guess, there seems to be nothing in the rules that prevents people from submitting guesses. In which case Doug is really asking people to pay $10 to enter what could be regarded as a lottery with a chance of winning $100000. Can Doug confirm that this competition satisfies all the legal requirements?
Anders, the details of the challenge are rather disturbingly vague so it's difficult to know just what is intended, but one thing that is clear is winning the money would require correctly identifying at least 900 of the 1,000 series as belonging to one group or the other. We don't know the distribution of series created via a trendless/trending process, but it's difficult to believe anyone could simply guess a correct answer.
If there are an equal number of each, it'd be like calling 900 of 1,000 coin flips correctly. It's theoretically possible, but that's clearly not the intended solution. It's also not one that is being encouraged in any way. I doubt the fact a problem could be solved by guessing is, by itself, enough to make it considered a lottery under the law.
I'm not a lawyer though, so I could be wrong.
...what could be regarded as a lottery with a chance of winning $100000.
Unregistered Commenter...and Then There's Physics
Ken, I think you'll find that if a thing is a contest of skill (even though the winner may actually turn out to have no skill, been drunk at the time, and picked the results with a pin) it is not a lottery.
Well bugger me sideways with a camel, Brandon. You've got it in one.
You, don't have to be a lawyer to see that Keenan is yanking their chain, saying "Come on then, if you think you're hard enough. I can play at your level, and win." ATTP isn't putting up his own money.
ATTp is almost, but not quite, at the level of Slartibartfast
@ Brandon Shollenberger
Regarding the length of the time series, the main temperature series relied upon by the IPCC, in its most-recent Assessment Report (AR5), begins in 1880 and is annual.
Nov 18, 2015 at 4:00 PM | ...and Then There's Physics
"In the wide range of models considered, the odds are invariably in favour of the enhanced greenhouse effect as an explanation of the observed warming over the last century."
IOW, if one assumes that the climate evolves according to our hypothesis, then the climate evolves according to our hypothesis. Meh.
Douglas J. Keenan:
I'm pretty sure that's not true, but I suppose it's also not really important for this challenge. This challenge can stand on its own merits. I do have to ask though, do people attempting this challenge get to have any idea of approximately how many series fall in each category? Normally a person might expect the distribution to be about 50/50, but the challenge doesn't actually say.
It would be really cheesy if it turns out there was only one series created with a trending process, or some other crazy answer like that.
"It would be really cheesy if it turns out there was only one series created with a trending process, or some other crazy answer like that."
And, upon what basis do you expect nature to refrain from "cheesiness"?
If the series started in 1880 it is evidently 135 years long -presumably this mean there are 135 data points to analyse - one for each year? If so it is at least as good a record as ice core record annual data, which are treated as being valid climate indicators. Are ATTP, Russell and Brandon S. saying they can't use ice core data to define climate trends?
/
Bart:
Say what? Nature isn't going to create 1,000 series and offer people $100,000 if they are able to categorize at least 900 of them correctly.
Spectator:
Doug Keenan is claiming "the main temperature series relied upon by the IPCC" is an annual series. I said I don't believe that is true. Nothing about that statement suggests anything like what you claim.
Due respect to Doug Keenan, it's worth the money just to see brandon, ken and russell getting their underwear in a twist.
Salopian, are you going to include me in that august company? I thought my comment was reasonable, as was part at least of ATTP's.
Tony.
And why am I being included in that list? What have I done to be grouped with Anders and Russell Seitz? I would think my primary concern is completely reasonable - whether or not there is any assurance the money would be paid if somebody met the requirements.
As for the other issues I raised, they seem rather reasonable to me. People do usually look at monthly temperatures, not annual ones, and it is reasonable to ask if we can be given an idea of how approximately many series we should expect in each category. A similar concern would be this challenge says the added trends have an "average" of over one degree, but that doesn't tell us how small they could be. If some of the trends are very small, on the scale of .05/century, it'd probably be impossible to detect them.
But asking questions about issues like these should hardly be taken as getting one's "underwear in a twist."
"Say what? Nature isn't going to create 1,000 series and offer people $100,000 if they are able to categorize at least 900 of them correctly."
Pas du tout. Nature has created a series. For you to establish that you have likely identified it correctly, you need to show that you have skill in performing such identifications without any a priori information.
Bart:
That's not true at all. When dealing with real temperatures, we have tons of a priori information. For instance, we know certain statistical models are simply impossible because there is no physical way for them to happen as they violate laws of physics. We also know far more than 135 data points because we can use unsmoothed data, rather than relying on annual data. So forth and so on.
Studying physical phenomena doesn't require us ignore all knowledge we have of the physical properties of what we're studying.
Tony: If you so wish, consider yourself included.
Brandon: No, I don't think your primary concern is in anyway reasonable - Doug Keenan has set a challenge, and you have just come back with some cheap shots to try to undermine it.
This is certainly going to be an interesting exercise. I have a strong doubt the challenge can be solved but since I am not a statistician there's a microscopic chance I will approach it in a way that means something. Probably not but it looks like fun.
I see many responders not understanding what exactly is the challenge. Out of 1000 sets of numbers, identify correctly at least 900 of them as to whether or not they have a trend. It has no declared connection to reality; it is a mathematical puzzle.
Yet the implication is obvious and that is what ATTP and others are reacting to; it casts a shadow of doubt on claims of certainty with regard to the "A" of AGW. Just a shadow of course; an uncertainty.
As people pursue this they will necessarily learn some fine points of math and statistics; how can that be a bad thing?
My sense right now, without even starting (I looked at the data), is that it would be trivial to guarantee no one wins the prize by simply ensuring that the trend is, on average, smaller than average variance over that short of a series.
I suspect a bit of aliasing is going to be present to create the appearance of trend where a trend doesn't actually exist.
Michael 2:
You have to be careful when saying this. It's not really whether or not the series have trends; it's whether those series had linear trends added after the fact. A series could have a trend because the noise model used to create it happened to come up with a trend in the noise it happened to create in that particular case.
The announcement does say each "trend averaged 1°C/century." That makes it sound (to me) like each series that had a trend added had the a trend of the same magnitude added, but I think what's meant is the average of all the trends added was 1°C/century. If that's true, we have no way to know how small some of the series could have had added to them.
On the other hand, if every series (that had a trend added) had a trend of the same magnitude added to it, this challenge will likely be pretty easy to beat.
Each trend averaged 1°C/century. In other words, if a series had a trend added to it, the trend was ±1°C/century, when averaged across the series. The trend added to a series was not necessarily deterministic-constant throughout the whole series though.
The wording has been changed slightly to make this clearer.
Ah, okay. Thanks for the clarification. I hadn't considered that possibility.
By the way, is there a method for submitting answers/payment yet?
Some many moons ago somebody, and I apologise for my lack of recall of the reference, likened our present day academic understanding of this planet and its climate as to be that of an adolescent teenager convinced they are already in receipt of everything it is ever possible to know.
Whilst the challenge set by Douglas J. Keenan is way above my comprehension, my reading of the majority of posts to this thread does appear to confirm my unnamed warriors observations.
Shame this show looks to have finished - I'd say this would make a good item for them:
Radio 4 "More or less"
http://www.bbc.co.uk/programmes/b006qshd
I was just goofing off a bit, and I came up with a guess at a list as an answer. I think that makes me the first?
I don't know what I'm supposed to do with it though. I mean, I'm not sure I'd want to spend $10 on a guess I put together so quickly, but even if I did, I don't see any way to submit our guesses.
Brandon
Chill, patience, enjoy waiting, like the great minds gone before we are destined never to know the whole!
Sit back, enjoy the ride, enjoy the knowledge, irrespective of derivation, or its illogical origination!
But most of all realise effectively we know nowt!
But, but, but at the end of OUR day somebody has to 'prove' we know nowt!
The games played by the ultra-rich baffle.
Green Sand, I don't get why you're telling me to have patience. A person announced a challenge. Naturally, people are going to want to know how they should submit answers for it. It appears I happen to be the first to come up with an answer, but that's not a matter of me lacking patience. It's just a matter of me coming up with a simple way to (try to) solve the challenge. I don't know that it would actually work, but sometimes simple solutions are the best.
Nov 18, 2015 at 10:18 PM | Brandon Shollenberger
"For instance, we know certain statistical models are simply impossible because there is no physical way for them to happen as they violate laws of physics."
You know that much about Keenan's series.
"We also know far more than 135 data points because we can use unsmoothed data, rather than relying on annual data."
Doesn't help a lot for diagnosing long term properties.
This is tiresome. Have the last word if you like.
It would be easier to break the encryption key.
The problem has always been that the latter-day Lysenkoists are trying to tease a very small putative signal out of extremely noisy data. Their methods, to date, have not been limited to ever more bizarre statistical contortions based on less and less reliable proxies, but have necessarily included blatantly outrageous "tricks."
Mr. Keenan's experiment/ challenge represents a clever reductio ad absurdum, one not vulnerable to tricks. On the one hand, if they refuse the challenge, they tacitly admit that their entire panoply of methods are useless for the aforementioned task. On the other hand, if they accept the challenge, they may be revealed as witch doctors. Is it better to remain silent and be thought fools? Or to speak up and remove all doubt?
Personally, I suspect that this challenge is may be subject to analysis via a simple method, which this margin is too narrow to contain.
@wottsy
I did not enter the competition, so I am not eligible for the price. You may enter the competition on my behalf, but the challenge is to correct identify at least 900 of a 1000 series; I am not sure that identifying a series that is not in the challenge-set, would qualify.
Really enjoyed this thread. Some did not get the nature of the challenge. Some that did, tried to divert it. Is pure and simple. Can you, or can you not, discern 'reliable' CAGW from among the samples provided, 1/3 of which mathematically preserve the hypothesized warming trend. Beautiful. My supposition, no one can. So the specific bet is merely a safe harbor. But boy, does it expose the climate charlitans.
Richard,
Ahhh, no :-) Yes, I read on after writing that first comment. I hadn't, initially, realised that the challenge was quite as silly as it actually is. Just when I thought that Doug couldn't possible illustrate his ignorance of this topic any more than he already has, he's goes ahead and surprises me. Possibly you - as someone who has published in this field - could point out to Doug that showing (or not) that we can identify time series that have had trends added to them, is entirely irrelevant when it comes to establishing if the increase in global mean surface temperature is due to random natural variation, or not. It is a potentially interesting challenge, but relevant to this issue it is not.
While I'm here, though, I'm now actually confused about what the challenge is saying. Does each of the 1000 time series have a linear trend of 1C/century, but some of them have had a trend added to them, and some have not? Or, as I first thought, there are 1000 time series, some of which have had a 1C/century trend added to them. Or, something else altogether.
Typical Wotts, writing comments on things he hasn't read.