This should give FirstAscent a stroke...

[myrkul]

(FYI, the intermediate tab has this gem: "On the other hand, Uranus is cooling")

Thought I felt a chill.

[scrybe]

Polynomial co-integration is a technique from the world of the Austrian School of Economics (http://www.ihs.ac.at/publications/eco/es-264.pdf) that is used when you are trying to find correlations between a bunch of data that you THINK is related, but you don't know HOW.

It's not just a matter of degree like you are stating, we know some stuff, not every little niggle, but we still know enough to show that the climate DOES have forcings (multiple, not just 1) that influence things from one year to the next. We know the basic rules (even if Bill O'Riley thinks we don't know what causes tides. Yes, he really said that.) and we know that the movement from year to year is not in a random pattern.

I'm pretty sure if I was a climate scientist I'd either be laughing or crying about how inappropriate this analysis method is. As it is I'm pretty appalled.

Look at a graph of temperature (even the average of average of average ones). Pretty much none of the "niggles" are explained, and are therefore validly modeled non-deterministically (ie stochastically). There may be problems with using cointegration to describe climate, but I am confident you are not describing them well since it makes no sense.

I'll admit that I'm only somewhat literate at higher math, let me try to frame it again and give you a couple links that should make it clear in a few minutes of reading.

Essentially they are using a root unit which should have been rejected because the CO2 dataset is not a random-walk. This analysis method is only for non-deterministic systems, and generally the data is tested to ensure that it is random and not deterministic. The researchers in this managed not to run any of the checks in an effective manner, and there is signal showing through where there should only be a trend + noise. This link has a blog post that gets into the math: http://tamino.wordpress.com/2010/03/16/still-not/ and this one has some explanations and expansions on it: http://citizenschallenge.blogspot.com/2013/01/polynomial-cointegration-tests-of.html (the second one is more readable, but you get a better sense of the math from the first)


Edit, shorter option from comments on original article:

Dr. Acula says:
January 3, 2013 at 12:26 pm
Being well-versed in the Austrian School of economics, I have a pretty low opinion of econometric techniques.

Sorry, but this paper seems to be playing mathematical games to me. It’s not at all obvious why “cointegration tests” should be trusted. What empirical evidence is there to justify using cointegration tests? Why were certain tests used and not their alternatives?

I’m guessing this isn’t really science, but rather the opinion of (perhaps seasoned) econometricists engaging in their art.

It did not take me long to find troubling information about cointegration: http://www.capco.com/capco-institute/capco-journal/journal-32-applied-finance/the-failure-of-financial-econometrics-asses

“This paper demonstrates that the results obtained by using different cointegration tests vary considerably and that they are not robust with respect to model specification. It is also demonstrated that, contrary to what is claimed, cointegration analysis does not allow distinction between spurious relations and genuine ones. Some of the pillars of cointegration analysis are not supported by the results presented in this study.”

AND
DeWitt Payne says:
January 3, 2013 at 2:15 pm

I will remind you that ALL of the top research departments of the world’s central banks use this methodology and result format.

And looking at the global economy, I would say it’s working really well. /sarc

Just one fundamental flaw of many. Atmospheric CO2 concentration is not a random variable. It is almost completely deterministic. There is measurement error and year to year variability, but those factors are small compared to the deterministic change. We know where it comes from and how much is emitted each year. Applying a unit root test to this data without removing the deterministic trend is therefore invalid.

(FYI, the intermediate tab has this gem: "On the other hand, Uranus is cooling")

Thought I felt a chill.

I might use that: Hey, Uranus is cooling, pull up your pants.

[Rassah]

How about addressing the data, instead of the source?

Ooh ooh ooh! Can I refute this on FirstAscent's behalf?

"I am not a climate scientist, and neither are you!"

How's that?

[bb113]

Polynomial co-integration is a technique from the world of the Austrian School of Economics (http://www.ihs.ac.at/publications/eco/es-264.pdf) that is used when you are trying to find correlations between a bunch of data that you THINK is related, but you don't know HOW.

It's not just a matter of degree like you are stating, we know some stuff, not every little niggle, but we still know enough to show that the climate DOES have forcings (multiple, not just 1) that influence things from one year to the next. We know the basic rules (even if Bill O'Riley thinks we don't know what causes tides. Yes, he really said that.) and we know that the movement from year to year is not in a random pattern.

I'm pretty sure if I was a climate scientist I'd either be laughing or crying about how inappropriate this analysis method is. As it is I'm pretty appalled.

Look at a graph of temperature (even the average of average of average ones). Pretty much none of the "niggles" are explained, and are therefore validly modeled non-deterministically (ie stochastically). There may be problems with using cointegration to describe climate, but I am confident you are not describing them well since it makes no sense.

I'll admit that I'm only somewhat literate at higher math, let me try to frame it again and give you a couple links that should make it clear in a few minutes of reading.

Essentially they are using a root unit which should have been rejected because the dataset is not a random-walk. This analysis method is only for non-deterministic systems, and generally the data is tested to ensure that it is random and not deterministic. The researchers in this managed not to run any of the checks in an effective manner, and there is signal showing through where there should only be a trend + noise. This link has a blog post that gets into the math: http://tamino.wordpress.com/2010/03/16/still-not/ and this one has some explanations and expansions on it: http://citizenschallenge.blogspot.com/2013/01/polynomial-cointegration-tests-of.html (the second one is more readable, but you get a better sense of the math from the first)


Edit, shorter option from comments on original article:

Dr. Acula says:
January 3, 2013 at 12:26 pm
Being well-versed in the Austrian School of economics, I have a pretty low opinion of econometric techniques.

Sorry, but this paper seems to be playing mathematical games to me. It’s not at all obvious why “cointegration tests” should be trusted. What empirical evidence is there to justify using cointegration tests? Why were certain tests used and not their alternatives?

I’m guessing this isn’t really science, but rather the opinion of (perhaps seasoned) econometricists engaging in their art.

It did not take me long to find troubling information about cointegration: http://www.capco.com/capco-institute/capco-journal/journal-32-applied-finance/the-failure-of-financial-econometrics-asses

“This paper demonstrates that the results obtained by using different cointegration tests vary considerably and that they are not robust with respect to model specification. It is also demonstrated that, contrary to what is claimed, cointegration analysis does not allow distinction between spurious relations and genuine ones. Some of the pillars of cointegration analysis are not supported by the results presented in this study.”

Well all I can say is these debunkers fail to do proper debunking as well (eg "I found some troubling info on conintegration techniques") . Just so you know you can find thousands of papers saying that about null hypothesis significance testing (most commonly used statistical strategy), so if you think finding a published paper criticizing a stats technique invalidates a paper you are in my camp of all science since 1950 is on shaky ground.

The "Still Not" link seemed the best but is debunking some other paper and all they guy/girl does is some half assed power analysis. I suppose what I would like to see is an actual power analysis... which I guess I could do myself.

[scrybe]

more on cointegration:

One aspect of the failure of financial econometrics is the use of cointegration analysis for financial decision making and policy analysis. This paper demonstrates that the results obtained by using different cointegration tests vary considerably and that they are not robust with respect to model specification. It is also demonstrated that, contrary to what is claimed, cointegration analysis does not allow distinction between spurious relations and genuine ones. Some of the pillars of cointegration analysis are not supported by the results presented in this study. Specifically it is shown that cointegration does not necessarily imply, or is implied by, a valid error correction representation and that causality is not necessarily present in at least one direction. More importantly, however, cointegration analysis does not lead to sound financial decisions, and a better job can be done by using simple correlation analysis.

...

The objective of this paper is to demonstrate that cointegration analysis, error correction modeling, and causality testing are misleading, confusing, and provide a tool for proving preconceived ideas and beliefs. More important, however, is the hazardous practice of using the results of cointegration analysis to guide policy and financial operations, including investment, financing, and hedging. With the help of examples on stock market integration and international parity conditions it will be demonstrated that cointegration analysis produces results that tell us nothing and that for practical purposes these results are useless at best and dangerous at worst.

[bb113]

more on cointegration:

One aspect of the failure of financial econometrics is the use of cointegration analysis for financial decision making and policy analysis. This paper demonstrates that the results obtained by using different cointegration tests vary considerably and that they are not robust with respect to model specification. It is also demonstrated that, contrary to what is claimed, cointegration analysis does not allow distinction between spurious relations and genuine ones. Some of the pillars of cointegration analysis are not supported by the results presented in this study. Specifically it is shown that cointegration does not necessarily imply, or is implied by, a valid error correction representation and that causality is not necessarily present in at least one direction. More importantly, however, cointegration analysis does not lead to sound financial decisions, and a better job can be done by using simple correlation analysis.

Thats fine. All I'm saying is that if that is enough evidence for you to dismiss the paper you should dismiss pretty much every journal article in every field except physics since 1950. Read some quotes on this page:
https://en.wikipedia.org/wiki/Statistical_hypothesis_testing#Controversy

Example:

    Null hypotheses of no difference are usually known to be false before the data are collected; when they are, their rejection or acceptance simply reflects the size of the sample and the power of the test, and is not a contribution to science.

[myrkul]

Thats fine. All I'm saying is that if that is enough evidence for you to dismiss the paper you should dismiss pretty much every journal article in every field except physics since 1950.

What about genetics? Didn't Watson & Crick publish in '53?

[bb113]

Thats fine. All I'm saying is that if that is enough evidence for you to dismiss the paper you should dismiss pretty much every journal article in every field except physics since 1950.

What about genetics? Didn't Watson & Crick publish in '53?

No stats in that paper...

 Look at this current paper and the source of the controversy, it really does prove my point. Previous work has attempted to reject the null hypothesis that there is no correlation between temperature and CO2. This current work tests the null hypothesis that there is no autocorrelation. Surprise surprise, everyone proves their hypothesis true since there is always some correlation and (probably) always some degree of autocorrelation in any timeseries data set.

The proper thing to do (which the climate scientists have done with their models) is predict something and see if the model fits, then explain the deviations and make the model better. The current paper predicts nothing as far as i could tell so to me its kinda dumb regardless of how valid the method is. The climate models don't seem to be predicting the temperature all that well, meaning they can be improved.

[myrkul]

Thats fine. All I'm saying is that if that is enough evidence for you to dismiss the paper you should dismiss pretty much every journal article in every field except physics since 1950.

What about genetics? Didn't Watson & Crick publish in '53?

No stats in that paper...

 Look at this current paper and the source of the controversy, it really does prove my point. Previous work has attempted to reject the null hypothesis that there is no correlation between temperature and CO2. This current work tests the null hypothesis that there is no autocorrelation. Surprise surprise, everyone proves their hypothesis true since there is always some correlation and (probably) always some degree of autocorrelation in any timeseries data set.

"Lies, damn lies, and statistics," eh?

[ElectricMucus]

I thought I'd post that for hilarity. Apparently there are SUVs on Neptune.
Oh and I researched the "debunking" too, the data on solar irradiance is supposed to be false. If it weren't for the Neptune chart that would be almost believable

Well hello there, I had to take you off ignore for this one (FYI, I found several of your comments in this thread insightful and thought provoking, I'm not sure what is wrong with me )

I'm pretty sure that the "other planets are warming too" thing came from a session where they just compared all the graphs to find several that went in the same direction.

Here is a good recap of the details: http://www.skepticalscience.com/global-warming-other-planets-solar-system-basic.htm

I like the Neptune one the best because most people have no concept of the speed seasons change with that 164 year orbital period. Not sure if they have SUV's or not, but if they have spring fashions like our coed's do, you might want to visit a Neptuntian University and take in the sights.

Also the earth solar forcing has been uncooperative since the temperature keeps going up when it goes up, or down, or stays the same.

(FYI, the intermediate tab has this gem: "On the other hand, Uranus is cooling")

Yes I've seen that too.

There are bunch of excuses for other planets in there too, all of them different. Not really that consistent.

[bb113]

Thats fine. All I'm saying is that if that is enough evidence for you to dismiss the paper you should dismiss pretty much every journal article in every field except physics since 1950.

What about genetics? Didn't Watson & Crick publish in '53?

No stats in that paper...

 Look at this current paper and the source of the controversy, it really does prove my point. Previous work has attempted to reject the null hypothesis that there is no correlation between temperature and CO2. This current work tests the null hypothesis that there is no autocorrelation. Surprise surprise, everyone proves their hypothesis true since there is always some correlation and (probably) always some degree of autocorrelation in any timeseries data set.

"Lies, damn lies, and statistics," eh?

I actually like statistics, its just almost no scientist receives proper training (me included) but then still feel that they are qualified to use stats and draw conclusions from their results.

[FirstAscent]

The climate models don't seem to be predicting the temperature all that well, meaning they can be improved.

You're right. They didn't predict that the Arctic ice would melt as fast as it is. But they have been consistently predicting it with an increasing consensus for forty or so years. Myrkul of course likes to pull one of his deniers' memes from his ass about how they were predicting a coming ice age in the '70s, but that's just the deniers picking up on the fact that there was a paper or two published by a few scientists back then, totally ignoring the fact that many more scientists were on board with global warming.

[bb113]

The climate models don't seem to be predicting the temperature all that well, meaning they can be improved.

You're right. They didn't predict that the Arctic ice would melt as fast as it is. But they have been consistently predicting it with an increasing consensus for forty or so years.

This sounds absurd to me. Who cares about consensus. The models predict something, we see if it occurs. Predicting the direction of something is 50/50 a priori. Is this the paper you are referring to?:

http://neven1.typepad.com/blog/2012/09/models-are-improving-but-can-they-catch-up.html

[scrybe]

Yes I've seen that too.

There are bunch of excuses for other planets in there too, all of them different. Not really that consistent.

Yeah, showing a 30 year graph when a season is 40 years long is not exactly the best idea. I'm sure most statisticians would freak out pretty badly when they realize you don't even have 1 complete sample, let alone enough to draw conclusions from.

If the planets were consistent this would be a boring universe. There really is no reason they SHOULD be consistent unless there really IS a warming effect from the SUN or other system-wide influence. It should also fall off by the square of the distance and have a lesser impact (ignoring albedo and a lot of other stuff, granted, but square laws add up over millions of miles) and be something that is consistent across all the planets, not just 8 of 100. There is also the minor detail of REDUCED solar output as the sun is going through a cooling trend ATM.

Calling these explanations "excuses" implies that you don't believe them, is that the case?

[scrybe]

The climate models don't seem to be predicting the temperature all that well, meaning they can be improved.

You're right. They didn't predict that the Arctic ice would melt as fast as it is. But they have been consistently predicting it with an increasing consensus for forty or so years.

This sounds absurd to me. Who cares about consensus. The models predict something, we see if it occurs. Predicting the direction of something is 50/50 a priori. Is this the paper you are referring to?:

http://neven1.typepad.com/blog/2012/09/models-are-improving-but-can-they-catch-up.html

We care about scientific consensus because it is an indicator that the layman can use to determine if a scientist is in the mainstream of his field or a whackjob on the fringe (or in another field entirely.) Even if they can't agree on all the particulars, many folks have been predicting additional melting, and some have not. Those that have not have been moved out of the consensus view over time and are now on the fringe, but in the 80's it was not as certain and they were part of the consensus that said "we don't know what it will look like in 30 years" until they started getting a better picture. Additionally we are never going to have 1 "perfect climate model" so running multiple models and combining the results is another type of consensus that might have been intended.

Unfortunately some folks refuse to believe that their pet theory could be wrong, and think that 98% of the scientists must have an agenda to destroy it. These folks a) do not understand scientific consensus and b) are fooling themselves if they think that thousands of academics CAN get along well enough for a conspiracy work, and not leak like a sieve.

Predicting the direction of something is 50/50 in a random-walk model, when did you prove this earlier? In a model with causation you generally can predict the direction and magnitude with some level of certainty (and you should be able to generate some error bars as well.)

Thanks for the link BTW, good one, it looks like some of the models have been updated more than others, I'm seeing a lot of concern that the IPCC might rely more heavily on the ones that are not predicting an ice-free arctic anytime soon, we'll have to see when the report comes out. Given the size and makeup of the body it actually is fairly conservative, even if some folks would have you think otherwise.

[ElectricMucus]

Calling these explanations "excuses" implies that you don't believe them, is that the case?

Exactly.

IMO we simply don't know enough about our planet and the universe to draw a conclusion such as global warming is caused by us or not. I think we cannot know it the same way as we cannot know how to cure cancer.
There isn't enough research (By that I mean real research not throwing publications at each other in order to "prove" ones point of view.)

But there won't be any real research as long as it is a political issue. This is just how the society works, sadly.

[scrybe]

Calling these explanations "excuses" implies that you don't believe them, is that the case?

Exactly.

IMO we simply don't know enough about our planet and the universe to draw a conclusion such as global warming is caused by us or not. I think we cannot know it the same way as we cannot know how to cure cancer.
There isn't enough research (By that I mean real research not throwing publications at each other in order to "prove" ones point of view.)

But there won't be any real research as long as it is a political issue. This is just how the society works, sadly.

Really? Cancer? You are going to use a disease that we have been battling for years and have made good progress on (>1.6% reduction in mortality every year since 1979 in the US) and actually understand a lot of things about? Childhood Leukemia has an 80% cure rate with low remission, and non-melanoma skin cancers are not even counted in most cancer incidence statistics anymore since they are easily cured on 2-3 million people every year.

I don't know WTF you mean by "real research" but I'm pretty sure that thousands of folks would give you a beat-down for discounting their lifelong contributions.

We understand a lot more about the universe than you seem to realize, maybe you should step outside of Plato's cave sometimes.

I guess I was wrong about taking you off of ignore. You and Bill O'Riley can go back to scratching your heads about why the tides happen.

[ElectricMucus]

Calling these explanations "excuses" implies that you don't believe them, is that the case?

Exactly.

IMO we simply don't know enough about our planet and the universe to draw a conclusion such as global warming is caused by us or not. I think we cannot know it the same way as we cannot know how to cure cancer.
There isn't enough research (By that I mean real research not throwing publications at each other in order to "prove" ones point of view.)

But there won't be any real research as long as it is a political issue. This is just how the society works, sadly.

Really? Cancer? You are going to use a disease that we have been battling for years and have made good progress on (>1.6% reduction in mortality every year since 1979 in the US) and actually understand a lot of things about? Childhood Leukemia has an 80% cure rate with low remission, and non-melanoma skin cancers are not even counted in most cancer incidence statistics anymore since they are easily cured on 2-3 million people every year.

We understand a lot more about the universe than you seem to realize, maybe you should step outside of Plato's cave sometimes.

I guess I was wrong about taking you off of ignore. You and Bill O'Riley can go back to scratching your heads about why the tides happen.

What it is always with threatening ignore with you people when you disagree on something I typed?

Maybe cancer wasn't the best example, my impression is that besides threating the symptoms and removing the affected areas of the body (which doesn't always work) there is not much we can do.
In the medieval ages they had treatments for "fever" too if I may bring up another analogy... But a cure NFW!

[scrybe]

Calling these explanations "excuses" implies that you don't believe them, is that the case?

Exactly.

IMO we simply don't know enough about our planet and the universe to draw a conclusion such as global warming is caused by us or not. I think we cannot know it the same way as we cannot know how to cure cancer.
There isn't enough research (By that I mean real research not throwing publications at each other in order to "prove" ones point of view.)

But there won't be any real research as long as it is a political issue. This is just how the society works, sadly.

Really? Cancer? You are going to use a disease that we have been battling for years and have made good progress on (>1.6% reduction in mortality every year since 1979 in the US) and actually understand a lot of things about? Childhood Leukemia has an 80% cure rate with low remission, and non-melanoma skin cancers are not even counted in most cancer incidence statistics anymore since they are easily cured on 2-3 million people every year.

We understand a lot more about the universe than you seem to realize, maybe you should step outside of Plato's cave sometimes.

I guess I was wrong about taking you off of ignore. You and Bill O'Riley can go back to scratching your heads about why the tides happen.

What it is always with threatening ignore with you people when you disagree on something I typed?

Maybe cancer wasn't the best example, my impression is that besides threating the symptoms and removing the affected areas of the body (which doesn't always work) there is not much we can do.
In the medieval ages they had treatments for "fever" too if I may bring up another analogy... But a cure NFW!

http://www.nytimes.com/2011/09/13/health/13gene.html?pagewanted=all&_r=0

We are getting a LOT closer than you think, I have a friend who used her own immune system to kill (metastasized) pancreatic cancer that would have been fatal 20 years ago.

Climate science is not new, we've been doing "real science" for decades. Same with gravitation, solar astronomy, thermodynamics, and a other disciplines that can give us a very good approximation of the answers, even if we don't understand all the details. All by itself the cherry-picking of 8 out of 100 orbital bodies should allow you to see the deception that is being attempted, there are a lot more examples.

Expecting politics to push science in the right direction is like expecting a pig to push the button on it's own bolt gun. Science generally FORCES our leaders to change by prooving that they are wrong (and in a democracy, getting enough constituents riled up,) but with the current crop of ideologues that refuse to give an inch on any point of dogma (no matter how much evidence is presented) we will be a long time waiting for some of them to come around.

I think you are either a troll or an idiot who cannot pay attention to the current state of the art and is willing to talk out of your ass to avoid being wrong. That is why I'm close putting you back on ignore, and people always threaten it because you act like a big fucking troll or an absolute idiot much of the time as your "deep yellow" link shows. If you have not figured it out by now, I most likely can't help, but I gave it a shot. Up until this conversation I had thought that this was a moderators troll-shill account, now I'm not sure, it might be the other option.

I said it earlier, this is not about "belief" it is about "scientific consensus" take a look at the literature and the VAST majority of it will fall in one direction, take a good hard look and the anti-science and pseudo-science stuff will start to look ridiculous and sometimes borderline delusional.

[ElectricMucus]

Oh maybe you really should put me on ignore since you are continuing to insult me. I have no problem with reading your insults, I just won't respond to them.