Monday, March 22, 2010

Tamino, 5 sigma, and frame-dragging

Grant Tamino Foster and other AGW fundamentalists can't swallow a bitter pill - that hard sciences still exist: see my previous text, Defending statistical methods.

Gravity Probe B, a largely failed project

He said that it was "ludicrous" to require 5 sigma confidence level as a criterion for the discovery of a new effect. (For non-professionals: the rule essentially says that the "signal" has to be at least 5 times bigger than the "noise" or "error" for you to be allowed to claim that you "see the signal" and it's not just a fluke.) I answered:
Dear Tamino,

five sigma is not a “ludicrously high” confidence level. In more serious disciplines of hard science, it’s the standard threshold for claiming a discovery of an effect. Search for “five sigma discovery” – you will find nearly 12,000 pages on Google:

Google: 5-sigma discovery (+2,000 spelled with "five")

It’s never ludicrous because the amount of data/events that you typically need for a 5-sigma discovery is just roughly 3 times bigger than the data for a 3-sigma discovery: the amount of required events often increases just as the square of the number of standard deviations because the relative error often goes like 1/sqrt(N).

So if a hypothesis is correct, it’s very likely that scientists will be able to do just a little bit of extra work to increase their 3-sigma claims to 5-sigma claims. If they can’t do it, it is really mostly because their original conjecture was wrong and the 3-sigma evidence was fake or a false positive (sometimes a cherry-picked, deliberately looked for one).

Cheers
Lubos
If you want to avoid technicalities and jargon, please skip the following paragraph.

You know, it's very unlikely for the evidence to "naturally" give you a 2-3 sigma evidence for an effect. Why? The "signal to noise ratio" is a priori distributed over several orders of magnitude - imagine the base-10 logarithm of the ratio is uniformly distributed between -3 and +3 - and for the ratio to be between 2 and 3 sigma, the logarithm has to be between 0.30 and 0.48. The width is just 0.18. If the a priori total span of the ratio were 6 orders of magnitude, the probability that the ratio leads to a 2-3 sigma evidence is just 0.18/6 = 3%. But papers claiming such 2-3 sigma findings are omnipresent - most of them must be wrong. It's much more likely for the evidence to show more than 5-sigma evidence than to be stuck between 2 and 3 sigmas. Also, we know why there are so many papers with wrong 2-3 sigma statements: people just cheat to get evidence for claims that they like (and that are mostly wrong, which is why they're so exciting).

See a nice reaction by Calabi, a physicist who got to Tamino's blog and defended 3-sigma and ideally 5-sigma results. Although Calabi, whoever (s)he actually is (Eugenio Calabi is 87, and a mathematician), may disagree with me in other contexts, be sure that there can't be any disagreement here. So Tamino is understandably as insulted by Calabi's comments as he is insulted by mine. ;-) So we may learn that the scientific standards as presented by Calabi are arrogant, parochial, and idiocy. :-)

As soon as Haelfix joined our team, the insults were enriched by "myopia", "outrageous", and even a comparison of ourselves to Penrose's theory of quantum consciousness - that analogy really hurts. ;-) Physics needs to take the lesson, we're told, and 5-sigma standards should be reduced to 2-sigma ones haha. These guys are insane nuts, weeds that got out of control and that must be cured with chemicals. They would be fired from every good enough or average regional college during their very first freshman exams - but they still manage to annoy people on the Internet and they actually control some institutions that are sometimes considered scientific.

A break for entertainment

Mission Impossible: in 2008, Phil Jones wrote Mike, Ray, and Caspar that "this message will self destruct in 10 seconds!" (see the end of the e-mail). An amusing intermezzo via Vítězslav Kremlík. Czech Made.

By the way, 10 days ago, Nathan Johnson, a student of atmospheric sciences in Arizona (nicknamed "Atmoz"), painted himself as "the author" and requested the Wikipedia page on Keith Briffa to be erased. I restored it: a reaction by William Connolley indicates that he was behind the censorship, too. Hat tip: papertiger

So after my comment, Tamino repeated it was ludicrous to "require" 5 sigma, claiming that it would be prohibitively expensive to achieve 5 sigma. I tried to explain him that he is completely wrong:
Dear Tamino,

once again, it’s not “ridiculous” to require a 5-sigma confidence as a condition for a serious scientist’s claim of a discovery of an effect. It’s what particle physicists, cosmologists, and many others require all the time. They require it from themselves and others, too. You would never see a serious team’s paper claiming a 3-sigma discovery of a new quark. Everyone knows it would be preposterous to make such far-reaching statements. In a serious discipline, 3 sigma is just a vague hint, not real evidence.

The purpose of statistics as a tool to fool the people may be to get “as much information as possible”. 
The word "edit" means that my text was censored by Tamino. The sentence he cut in the middle said that the purpose of science as a tool to cheat might be to get "as much information as possible" but the purpose of statistics as a tool of science was to get "as much correct information as possible, while avoiding wrong information almost all the time". Of course, it was already too inconvenient a fact.

He completely erased the second part of my posting that calculated how much you need to increase the number of events - or samples - to improve e.g. a 2-sigma result to a 5-sigma result, assuming that your hypothesis is actually valid. (I originally used 3 sigma, but let me offer you an even more realistic comparison because many really soft people are actually satisfied with 2 sigma.) Because the relative errors tend to decrease as "1/sqrt(N)" with the number of measurements, you need to increase "N" by the factor of "(5/2)^2 = 6.25 - just six times.

When you do so, the risk of a false positive decreases from 4% to less than one part per million! This is surely a good investment, isn't it? That's why it's normal in proper disciplines to demand 5 sigma for a discovery. What did Tamino reply?
[Response: How pathetic that you again focus on vague implications of fraud to justify your idiotic standard for statistics. Maybe that *is* the way you do science. I find your attitude disgusting. I think we've heard all from you we need to.]
A very constructive reply that focuses on the content, indeed! ;-) Note that there was no indication of any "fraud" in that particular comment of mine. But you know, they have heard enough to determine that I was a heretic. :-)

Tamino is one of those hysterical aggressive religious fanatics who should have been shot by the U.S. soldiers right after September 2001 when the war on terrorism began. He just can't swallow the fact that what he defends is an inferior discipline full of unjustified claims, poor standards, rubbish, and, yes, fraud - and that there actually exist disciplines that require a 100,000 times smaller false positive rate.

Another bigot called Wyman told me:
What I learned today: Curing AIDS isn’t serious science.
Well, of course that the 2-sigma subset of it is not. All discoveries of a cure for AIDS that have been supported by 2 sigma or 3 sigma evidence have been bogus. That's why we don't have a real cure today. That's why it isn't serious science. Be sure that when a drug works, you can easily show that it works by more than 5 sigma evidence. Most of the contemporary science about the health effects of A on B is scientific flapdoodle, a sophisticated version of homeopathy.

Steve L, another reader, seems to realize that serious fields of science require much higher confidence level. But he claims that it's important that all disciplines can choose whatever standards they like. Well, they can, but when they choose very poor standards and flood their journals with bogus claims, no one will protect them from the valid statements that their discipline sucks and most of the researchers who pursue it are just pieces of dirty garbage which is the case of the environmental sciences and a few others.

Ray Ladbury also admits that that particle physicists require 5 sigma. But he claims that this extra stringent condition is needed for them because the researchers would otherwise go "bump hunting" which is bad. Apparently, the people in climate science are so neutral that they would never, ever, go bump hunting ;-) so 3 sigma or often 2 sigma is just enough in the context of these saints! :-) Moreover, they usually go "blade hunting" and prefer to hide the bumps which is OK.

Jonathan Gilligan has yet another argument to claim that I am doing a mistake:
Lubos misses the important matter of distinguishing the practical significance (e.g., clinical significance) from statistical significance. One way of looking at this is to consider the cost of a Type-I error and the cost of a Type-II error. In quark-hunting, the cost of a Type-I error is high (incorrectly reporting a new discovery), while a Type-II error is much less costly (delaying the correct report of a new discovery). I am treating delays as potential Type-II errors because delaying action is a decision to act, at least temporarily, as though the null hypothesis is true; thus, delaying action while waiting for further tests can, in practice, be a type-II error if knowing the true state of things would compel prompt action.

Thus, it pays quark hunters to wait until they’re very certain....
Oh, great, so in particle physics, it would be a real disaster to report a false positive - a new particle that doesn't exist. Billions of people would apparently die if they were told about a new quark that isn't there. (Well, 6.8 billion people in the world don't give a damn about new quarks - and I was shocked to learn that there even exist people on this planet who don't fully understand the importance of string theory haha.) But in climate science, it doesn't hurt to report a new global warming threat. One only pays a few trillion dollars per decade which is pretty much zero.

A similar, albeit less extreme, situation appears in the science that looks for cures for various diseases. A false positive means that billions of dollars are being paid for drugs that don't work - and may have bad side-effects, while patients and potential patients are led to preemptively restrict their lives because of reasons that are superstitions.

Clearly, the fact that higher standards are expected in particle physics has nothing to do with the real costs of the false positives - which would be zero (for the society). It has everything to do with the quality of the field, the researchers, and their scientific integrity, impartiality, and ability to resist corruption and politicization.

Another user, stewart, says that it's OK to have a poor confidence level because even with a poor confidence level, they can do a more "convincing" case to a sponsor and get the funding from him, especially if they repeat the same low-confidence result many times (the Goebbelsian method). Stewart, aren't you worried that what you're doing is called fraud?

Gravity Probe B: a failed experiment

Finally, Horatio Algeranon, one more AGW bigot, shoots another bullet:
Wonder what Motl thinks of the Gravity Probe B Frame dragging result
Our lastest data analysis indicates clear observation of frame-dragging. The statistical uncertainty is 14% (~5 marcs/yr)."
Or of this earlier result “Confirming the Frame-Dragging Effect with Satellite Laser Ranging”
the Satellite laser tracking to LAGEOS-1 and -2 appears to confirm General Relativity’s prediction of the Lense-Thirring precession at the 8-12% level (1-sigma)
But hey, it’s always a piece of cake for scientists to confirm this stuff to 5-sigma (or better) — and anything less is just useless fluff.
Well, that's very easy to say what "I" think about it. Gravity Probe B hasn't yet experimentally decided on the question whether frame-dragging exists or not. Of course that it hasn't. And it probably never will. It's not just my opinion: everyone agrees with that. The presentation on "confirming" wasn't an announcement of their result but just a plan for the project. But the result on this main effect hasn't actually arrived yet. Two years ago, we announced that frame-dragging was just twice smaller than the resolution of Gravity Probe B.

The situation has improved a bit but they still haven't claimed the observation. For a few years, it has seemed very likely that the low inherent sensitivity of the probe (an extra source of errors which was previously not appreciated) is actually an obstacle that will prevent them from ever observing the effect - and the whole \$0.7 billion project has been a failure. Yes, we hadn't known it before because the experiment has looked as a great idea - but now we do know that the money has been largely wasted for "useless fluff".

A NASA 2008 review has chosen Gravity Probe B to be the only one among ten missions that was classified as "failure" and received no new funding. (Yes, it was the dead last.) Of course, conspiracy theorists interpret the result as a "negative result" concerning frame-dragging. The real answer is that because of its technical limitations, the probe couldn't have told us either way - which is a different conclusion. (And believe me as a theorist that frame-dragging surely does exist even if we incapable of measuring it now.)

I am sure that the scientific weeds similar to the AGW types would have claimed a discovery many years ago - and they would steal billions of additional dollars from the taxpayers for a probe that can't really probe the desired effect (and they would claim that 1 or 2 sigma is so nice and consistent with the theory!) - but real scientists simply don't do such things because it would be highly unethical and unscientific.

Other complaints

I will omit a dirty scumbag who attacks me purely personally. But another bigot, Ed Darrell, wrote:
Reading Stewart’s post, I gather that Lubos won’t grant credence to anything that shows human causation or warming until we burn up three or four more planets to be sure we understand the effect, right?

What dimension does Lubos live in? We only have one Earth.
Well, I personally live in all dimensions of the world (unless the Standard Model is confined on a brane) - and there are probably 9+1 or 10+1 of them, whether Mr Darrell (or climate skeptics!) like it or not. Also, I live in a galaxy with hundreds of billions of stars (and many planets), in a universe with hundreds of billion of galaxies. That's very different from Darrell's universe with one planet.

You don't have to burn five planets but you still do need a five-sigma evidence at least on one of them to scientifically claim a discovery of a new effect.

Big claims about your magic ability to burn many planets at once can't replace the old-fashioned five-sigma scientific evidence. The statements about your power to burn five planets may be effectively used during your pissing contest with Al Gore, James Hansen, and other retarded boys from the kindergarten - but they don't work in science. (On Saturday, Bill Clinton made a great comment about Al Gore in D.C. He noted he was speaking on the night before the start of spring, “otherwise known to Al Gore as proof of global warming.”)

In the case of a dangerous warming, five-sigma evidence is much more modest a condition than geoengineering on 5 planets - and it's one that is not available at this time which surely does indicate something.

Summary

The AGW proponents are immoral scum and it's important to at least localize the contagious disease.

And that's the memo.