Comments on: Evidence (and global warming)

By: Nabakov

Nabakov — Thu, 09 Oct 2003 11:43:46 +0000

The very length of this thread (shurely shome record): highlights Mark Twain’s comment “Everyone talks about the weather but no one ever does anything about it”.

By: dsquared

dsquared — Wed, 08 Oct 2003 18:04:48 +0000

If I didn’t occasionally monstrously overestimate the importance of marginal results and apply them to non-marginal contexts, I think they’d take my economists’ licence away …

By: Jack

Jack — Wed, 08 Oct 2003 17:12:28 +0000

d-squared—I wasn’t really getting at something as subtle as a system with multiple attractors. I just wanted to point out that Chris and Prof. Stott say may start out with such different estimations of the probablities involved that even if they pay scrupulous attention to the event of record temperatures, the same single event may not bring their views closer together. For over elaborate example if someone believed in a sunspot driven cycle they might even violate Chris’ tacit hypothesis andin any case they will likely judge recent events similar events as not that likely. Combined with an overwhelming initial probability the result might shift rather less than those of someone initially undecided who thinks that a hot summer is far more likely given global warming than not. The enviroskeptic may actually have a probability more different from undecided’s after a single iteration, even if both get pushed in the same direction. With repeated samples they will both converge to the same result (ergodicity assumed etc.) but the convinced skeptic may well be a slow learner, even if they do accept some kind of Bayesian imperative. Reading some apologists for the Whitehouse over the Plame affair, I’m not sure that’s a mild assumption.

By: dsquared

dsquared — Wed, 08 Oct 2003 16:11:09 +0000

Just to clear up, I was referring to “Bayesian statistics”, which is a bit of a bastard child of Bayesian probability theory, being the branch of mathematics aimed at making tractable the estimation of statistical parameters of interest through iterative use of Bayes’ rule. Which does assume exchangeability in this way. Although note above, Chris was stating ex hypothesi that P(E|S1) > P(E|S2)

By: S. Baum

S. Baum — Wed, 08 Oct 2003 16:03:31 +0000

The theory being advanced by some as evidence of the sinking of "the Good Ships 'Global Warming' and 'Kyoto Protocol'", i.e. the ostensible refutation of the theory that anthropogenic increases in atmospheric CO2 concentration over the last hundred years are causing the measured increases in global average temperature, is well summarized in The Milky Way's Spiral Arms and Ice Age Epochs on Earth on Nir Shaviv's home page. The first paragraph of Shaviv's summary reads:

There are indications that Cosmic Rays affect climate on Earth. If this is true, then one should expect climatic variations while we roam the galaxy. This is because the density of cosmic ray sources in the galaxy is not uniform. In fact, it is concentrated in the galactic spiral arms (it arises from supernovae, which in our galaxy arise predominantly from the death of massive stars, which in turn form and die predominantly in spiral arms). Thus, each time we cross a galactic arm, we should expect a colder climate. Current data for the spiral arm passages gives a crossing once every 135+/-25 Million years.

That is, he's discovered a plausible and interesting quasi-periodic mechanism for explaining changes in global average temperature, with a periodicity of 135+/-25 million years. Inspection of Figure 2 of his summary gives a generous estimate of a maximum change of 5 deg. C over 100 million years, or 0.00000005 deg./yr. If we look at IPCC Fig. 5.16, we see that the measured global average temperature change over the last 140 years is, being conservative, about 0.7 deg. C, or 0.005 deg./yr.

Those interested (in the good way, of course) in the scientific debate might want to attempt to explain how a mechanism that changes global temperatures at a rate 1 million times more slowly than has been observed over the last 140 years can be used to supposedly "sink" another proposed mechanism that predicts that same order of magnitude of yearly temperature change that is observed.

Shaviv's work provides another piece that will undoubtedly help to solve several puzzling questions about the climate during the Phanerozoic, but the million-fold difference in time scales doesn't make it a realistic candidate for explaining global temperature increases over the last century.

By: Brian Weatherson

Brian Weatherson — Wed, 08 Oct 2003 15:45:01 +0000

I thought under a really subjective Bayesianism, you could have any kind of divergence you wanted under any conditions. Most of the convergence results I’ve seen assume that all parties have some particular attitude towards the possible pieces of evidence. (Exchangability is the most common assumption.) But there’s nothing in Bayesianism per se that requires exchangability. (I think the more complicated results dsquared mentions are cases where you make assumptions like exchangability and you still don’t get convergence. But I could be misremembering.) To put the point another way, for any logically independent H and E, and any x in [0,1] Pr(H|E) can be x consistent with Bayesian principles, so any amount of divergence with respect to H is possible after any evidence E.

By: dsquared

dsquared — Wed, 08 Oct 2003 15:03:04 +0000

Jack: Nonconvergence of Bayesian learners isn’t an issue when the two hypotheses are A and ‘A, as they are here. If it’s just a matter of a piece of evidence which is more likely under A than not-A, then convergence is strictly monotonic, IIRC (which I may not). You need more complicated decisions to get the local nonconvergence result.

By: Ray

Ray — Wed, 08 Oct 2003 11:43:56 +0000

I think, with regard to Bayesian aspects, the important thing here is the Bayes factor: the conditional probability of the observed datum given the hypothesis is true divided by the conditional probability of the observed datum given that the hypothesis is false. If the Bayes factor is sufficiently small, I think, for the purposes of human beings with heuristic minds, it’s safer to call something “no evidence” than evidence. I believe the term of the art is “statistically insignificant,” much as we’d regard a coin flipping heads 5 times in a row as unusual, but not a sign that the coin is asymmetric in any serious way.

By: Ian

Ian — Wed, 08 Oct 2003 10:40:59 +0000

Are there not two issues in the climate change debate? First is climate change actually taking place? Second – if it is what are the causes? The evidence appears to be stacking up in favour of a yes to the first question but the answer to the second is much less clear cut. A third question then arises – assuming the answer to Q1 is yes and to Q2 is don’t know – what do we do as a race? I think there are one or two people in Bangladesh (and Holland and East Anglia and …) who might be interested in an answer

By: Jack

Jack — Wed, 08 Oct 2003 07:44:13 +0000

I think it is interesting to discover from exploring Chris’s point that, because Bayesian learning depends upon initial assumptions, the same piece of information might not draw two bayesian learners closer together, even if they are not disingenuous, at least in the short run. I also wonder if people really do engage in Bayesian learing on this sort of issue. It actually seems more plausible that they engage in some kind of dynamic Lakatosian (urk, is that a word?) monster barring exercise. That is instead of deciding between a number of static theories in the light of subsequent events and initial assumptions, they instead recalibrate and rebuild their theory to make it work with the known history of events as a central proposition. If that is normal human operating procedure, what does prof Stott have to do to demonstrate the good faith necessary for dialogue? I think it unlikely that people carry fixed or accurate probabilities for certain events around in their heads with any degree of reliability. In particular people are quite bad at tweaking their understanding—big changes stick but marginal ones are hard to track.

By: mgl

mgl — Wed, 08 Oct 2003 06:39:07 +0000

Sorry, s. baum. I thought you were responding to my post. It seems you weren’t. On this: …discussing motives is boring and pointless vis a vis whether or not anthropogenic contributions are indeed warming the planet. …I couldn’t agree more. Thanks.

By: msg

msg — Wed, 08 Oct 2003 06:11:45 +0000

The levels of erudition and mannered discourse here, though fluctuating, are still higher than any I’m used to elsewhere. Which makes me feel like Buddy Hackett somehow. But the topic is one that a lot of us are vitally concerned with, some to the point of despair. There is much more than a trickle of first hand evidence of major climate shift, or change, it’s an avalanche now, even through the fog of backspin and intentionally planted hysteria you can hear it. The circumpolar people were talking about it a few years back, hunters dependent on ice holding, on animals being in a certain phase of their cycle, finding things coming earlier than ever, or later. Though recently I haven’t been hearing much from up there. Wasn’t the Matterhorn closed to climbing this summer because of the danger of thawing and consequent rock fall? Didn’t some reputable journal state unequivocally that every glacier on the planet is shrinking, and more rapidly than previously thought? I’m suggesting that the people on the front lines of this subject, the field researchers themselves, are being driven to hysteria or numb apathy by their findings. I’m humbly suggesting that the possiblity of climate change on the scale it’s proposed, and what it implies, is so enormous, and horrifying, that it’s causing neurotic reactions in almost all of us. Denial and an excessive interest in the dissection of minutiae, the pathologies of shock. And I’d like to suggest that intellectual prowess is a lot like wealth, an equivalent responsibility attends its gifts.

By: s. baum

s. baum — Wed, 08 Oct 2003 05:57:11 +0000

Mum always said that intemperance and sarcasm would be my downfall. I guess I’ll have to retract all those climate research papers now that my betters have taken the bloody piss out of me. Funny thing is, I don’t get intemperate and sarcastic until somebody impugns my motives, e.g.

“But climatic science doesn’t seem to be strong enough to do the job, and the anti warming folks are perfectly right to point out that many of the scientists are interested parties.”

That’s about as empty a statement as one can make, even if one cared to offer an operational definition of “interested” that didn’t cut at least equally both ways. Or, to put it another way, discussing motives is boring and pointless vis a vis whether or not anthropogenic contributions are indeed warming the planet. On the other hand, topics such as the statistics of extreme events, the parallelization of finite difference (and other) numerical methods, the parameterization of turbulent processes, the modeling of radiation processes in the atmosphere, etc. are both interesting and very much to the point. On yet another hand, if that’s still too intemperate and sarcastic for the local delicate (as opposed to climate) sensitivities, then I’ll get back to a nasty PDL script whose last (hopefully) nasty bug still eludes me and leave the important stuff to the medieval scholars.

By: mgl

mgl — Wed, 08 Oct 2003 05:20:00 +0000

Thanks for illustrating my point, bigring55t.

By: bigring55t

bigring55t — Wed, 08 Oct 2003 04:36:21 +0000

To be succinct, the reason to question the motivations of enviro skeptics is the same reason one would question the motivations of those scientists that work for the tobacco companies and continue to produce “research” that shows smoking isn’t deleterious to your health.