Bishop Hill Thingummydoodle noodle
Brandon Shollenberger has a lovely post up looking at some recent comments by Skeptical Science insider Tom Curtis and Anders Thingummydoodle from the "And Then There's Physics" blog. Readers will remember Anders as the chap who berated me about one of my posts on Doug Keenan's work, saying that it was a physical model you needed in order to understand what was causing global warming. This despite my having said almost precisely that in the blog post.
Anyway, Anders has been sounding off about the Hockey Stick, accusing McIntyre and McKitrick of all manner of sins and demonstrating in the process that he has absolutely no idea of how Mann got from his raw data through to his final reconstruction. His allegations are therefore completely and utterly wrong.
This paper is five pages long. It is completely impossible anyone would look at it and think ten cherry-picked graphs were included in it. It might take all of thirty seconds to scroll through the paper and see such graphs aren’t in it.
And Tom Curtis's defence is little better
Curtis refers to the simulations McIntyre ran as creating “pseudo reconstructions.” That is wrong. These are pseudo proxies. Reconstructions are created by combining multiple proxies. Mixing such things up suggests a poor grasp of the issues at hand.
Bit of a car crash really. But read the whole thing. These guys are climate's very own Keystone Kops.
Reader Comments (64)
If only there was a nice, well-written, easily understandable explanation in book form which they could read.
Anders' is where the loonies congress with each other and I would add the commenting skeptics to that list. It shows what years of glorified cartoonists as science communicators degenerates into. Leave the place alone just like you would of a white supremacist event.
There be monsters.
"glorified cartoonists as science communicators"
IIRC, Cook used to use cartoons to communicate religion. Which seems apposite.
I probably shouldn't bother, but just to be clear (and since you're obviously such an honest chap, I assume you like things to be clear), the incident in question happened on 12/12/2013 and I did indeed confuse the number of figures in the Wegman report with the number of figures in McIntyre & McKitrick's 2005 GRL paper. I believe I have acknowledge this at least once, if not more than once. I no longer quite know what to do about that, other than invent a time machine, go back to 12/12/2013 and make sure I don't write the offending comment in the first place.
The only reason that this has reared it's head now is because someone appears to be upset that they're no longer welcome on my blog and - for reasons I haven't worked out and probably won't bother to establish - brought this up on Judith's blog (actually, I think it was meant as an illustration of my lack of integrity).
I don't actually think I've accused McIntyre & McKitrick of any sins (unless by sin you mean that some aspects of their paper are not as clearly explained as one might hope). If I have and you can point out where, I'd happily clarify or retract. All I think I've done (now that I've clarified my confusion regarding the MM05 paper and the Wegman report) is discuss something that I think is accepted as true - all the known comparisons between the results in MM05 (or from the MM05 analysis) and MBH98 are from a sample of 100, selected for being most hockey-stick-like.
Also, as regards Doug Keenan, if you do indeed agree that one needs a physical model, then presumably you also agree that most of what has been concluded by Doug Keenan is wrong or, if not wrong, irrelevant. Maybe we shouldn't start that all over again, though.
Anyway, I shall leave it at that.
And Anders is obviously in need of help:
the key quote from Brandon's post (admittedly about Curtis rather than Andrs, but Anders still needs the help)
"2) He repeatedly acts as though people should only care about the 100 cherry picked series which weren’t used in the paper. He focuses almost entirely upon those 100 unused series rather than looking at the 10,000 series the paper actually uses.
It’s a pretty remarkable way to deny what should be a truth we can all agree on. Michael Mann’s methodology is biased to mine for hockey sticks. Nobody can seriously argue otherwise. Anyone who denies it, a decade after it was made obvious, is a denier."
@rhoda,Aug 6, 2014 at 6:09 PM: An easily understandable explanation would not help at all. THEY decide what to read and what not to read. Actually, they are already deciding what I should read and what I should not read - they have successfully redefined a peer review process.
Diogenes,
I know I should probably just give up, but maybe you could clarify in what way the 100 weren't used in the paper. The paper (Figure 1) shows a comparison between a simulated series and MBH98 Hockey Stick Shaped Series. The simulated series is described as
However, the simulated series is one of those in the sample of 100 that is presented in the Supplementary Information to the paper. The sample of 100 is described in the paper as
without the paper mentioning that these 100 were not randomly selected from the 10000 simulations, but were selected on the basis of their Hockey Stick Index (HSI) and were selected as the 100 with the highest HSI's.
So, the only comparison with MBH98 in the paper is with a sample that happens to lie in the 100 most hockey stick like (out of 10000) and the only other comparisons I've ever seen (Wegman report) are also from this sample of 100. It could simply be chance, but then the only comparisons that have ever been made have been with the 1% of the full sample that are most like MBH98 Hockey Sticks.
Having said that, if you read MM05 carefully, this becomes reasonably clear (although you do need to think carefully about Figure 2), although they could have made it clearer that their sample of 100 that they present in the Supplementary Information was not randomly selected, but was selected to be most hockey stick-like.
I believe that there is some truth to this, so I don't deny that there the decentered method can amplify hockey sticks. However, and I'll admit that this is the bit I'm still trying to work out, the issue relates to how you center the timeseries. MBH98 centered them on the mean of some period in the 20th century, rather than the entire time period. As I understand it, if there is no trend, then this should not matter (since the mean should be the same). Therefore, even if it can amplify Hockey Sticks, it can't create them if they're not there in the first place (as I understand it at least).
I've also seen plenty of evidence to suggest that the centering issue doesn't really change the result and nothing definitive to show that it does. You can also go and look at recent papers that use different methods and different proxies and - broadly speaking - the results look the same. In a sense, I don't see either MBH98 or MM05 as all that relevent and should probably have stuck with my original goal, which was "don't talk about Hockey Sticks".
Anyway, I shall carry on with what I was doing before (watching a movie with my son) and leave it at that.
diogenes, to be fair, it may be possible the auxiliary information deserves criticisms. Because the link to it in the paper doesn't work anymore, I can't see how they described the data in it. I can't tell if a non-random selection was presented in a way that implied it is random.
But that's a minor issue. Even if the auxiliary information had issues, the paper itself did not. The paper cannot sensibly be criticized on this issue. The paper's results cannot be undermined by this issue.
Anyone who made any effort to genuinely understand this topic would see that. The only way people can keep saying these things is by willfully avoiding simple truths.
"Keystone cops" is spot-on for this clown and his fan club. Many of his posts consists of giving his opinion on things he clearly doesn't understand or hasn't even read, which makes good slapstick comedy. Months ago he claimed that MM05 had numerous easily explained issues, but has not yet got round to saying what these are.
"almost precisely" . What is that Bishop?
Guys, you have to read Ander's comment just above. It's priceless. He expresses his lack of concern regarding Michael Mann's methodology by saying:
Which is unquestionably true. Nobody has claimed Mann's methodology can produce hockey sticks out of nothing. What people say is it mines for hockey sticks, or that it cherry picks hockey sticks. Anders recognizes that it does this, but doesn't seem bothered, yet he also says:
Michael Mann used a methodology which inherently cherry-picked data, and he hid that fact when publishing results built upon it. Anders isn't very bothered by that. MM05 (supposedly) weren't clear in their SI their sample not used in their paper was cherry picked, and Anders thinks that's a serious problem he needs to criticize the paper for.
Um, yeah. Here's a thought Anders. If you think cherry picking data when giving examples you don't claim are representative is bad, cherry picking data as a key part of the methodology you use to produce your central results is certainly bad.
Anders, why do you still insist on crashing the car?
Perhaps a better area of enquiry would be why Mann's latest reconstructions are not hockey-stick shaped and why he has not made any comments on why his earlier papers were junk?
To your knowledge, has Mann ever reconciled the totally different shapes or explained why he had to break the hockey-stick in his later reconstructions, or what the implications might be?
NB that graph was made by commentator Carrick, a regular on Lucia's Blackboard and other sites
""almost precisely" . What is that Bishop?"
That's where the Bishop claimed to be in agreement, having said: "If you want to understand why temperature has changed, you will indeed need a physical model." and Anders replied: "No, I really don't think we are. If you want to understand GW you need a physical model."
http://www.bishop-hill.net/blog/2014/7/3/where-there-is-harmony-let-us-create-discord.html
It was a fun debate. :-)
Anders
I'm a bit surprised that you haven't worked out decentering yet, almost 9 years after the issue was very clearly explained by McIntyre and Mckitrick and 16 years after the original offence. We all know that short centering (which is not PCA as that is ordinarly understood) selectively picks out series which are trending up or down at the end of the series. That means short centering will pick out red noise series which just happen to be away from their long run values at the endpoint. The pseudo first "principal component" produced by short centering will be unduly affected by these red noise series and will therefore likely be hockey stick shaped even if the correct first principal component is not. If you want to call these red noise series hockey sticks that's fine and you are correct that if there are no series whose end point average is very different from its whole period average then short centering makes little difference. What everyone who has looked as this seriously knows (including Michael Mann, though he spins it rater differently) is that if principal components is done properly the hockey stick shapes don't appear till the third or fourthh principal component. So in the Mann case, where only the firs component was used in the regression stage, it does make a big difference. Of course if you include enough principal components in the regression stage so that you include the hockey stick shape then you can get Mann type results. But then the question is whether the hockey stick shapes are a good representation of the temperature field, as Mann claims. Arithmetically the amount of tree ring variance explained by the fourth principal component properly calculated is quite small, i.e the fourth princiapl component does not do a good job of representing temperature. And more importantly the hockey stick shape of the fourth princaipla component is almost entirely explained by a few problematic bristlecone pine series whose anomalous twentieth century growth was attributed by the original data gatherers to carbon fertilsation and tentatively by Steve McIntyre to the effects of tree response to bark damage. Whatever the reason far from the fourth principal component derived from the tree ring proxies reflecting a widespread pattern across all trees in the Northern hemisphere it reflects strange growth in a small area of the US in a particular tree species. This has all been well established since 2005. And its quite straightforward. Let's agree on this and then we can look at other studies.
It's just a cult, no different from parents with autistic kids latching onto fake cures. But they are the autistic kids instead of the parents, so chaos reigns.
Steve was able to demonstrate that the Mannian technique could generate a hockey stick from the phone book that should not be possible
"Thingummydoodle"
Is this his real surname? Sorry, I've been away for a while. I've never heard of this Thingummydoodle dude. Now I'll never forget him even if I wanted to.
Face meet Palm. Chuckle on my way to bed.
The point about decentering and mining hockey sticks is that red noise produces a lot more hockey stick like shapes than white noise does. It has to do with the fact that red noise has a lot larger low frequency components than e.g. white noise does. So you get big excursions,and some of them can resemble the signal (hockey stick from temperature) that you're use to correlate (aka "cherry picking") against.
I think Anders would find the Hockey Stick (which was a great read btw) to be a useful reference. Sadly he will never read it, nor is he likely to take the time that it would take to learn the issues on his own, so ignorance will continue to reign supreme on that front.
Brandon: "Because the link to it [supplementary information] in the paper [MM05] doesn't work anymore, I can't see how they described the data in it.
The supplementary information is available here. The "readme" file describes the 100 hockey sticks thus:
No indication as to how they were selected, so I agree that should have been made clearer. The "hockey stick index" values for the series are all above 1.9, and from the histogram in figure 2 these are the highest (or nearly so). However, the same histogram shows that very few series -- and these all come from trendless noise, remember -- have a hockey stick index below 1.25 (in absolute value); most are above 1.5. I think a more representative sample would have been equally convincing.
HaroldW, thanks. An additional quote from that file describing the data says:
Neither quotes says anything about how the examples were chosen, so a person could easily assume they were randomly chosen. That's wrong. If you select a non-representative sample, you should be clear the sample is non-representative. The paper was when it used an example. The SI should have done the same. That it didn't is a ding against the SI.
Of course, a minor lack of clarity like that is just that: minor. The paper is unaffected by this, and the code is easy to understand. These examples were the least important part of the SI, which was already less important than the paper itself. It's interesting these people would make such a big deal out of such a minor issue.
This is especially true since MBH is far worse when it comes to disclosure. It failed to describe important aspects of its methodology, and it even made false claims about its methodology (it claimed to use standard PCA, something we all know is false). For years, the authors refused to release any code, and to this day, they haven't released parts of it.
Of course, that doesn't bother Anders, Tom Curtis or willard. You won't see any of them writing blog posts or lengthy comments criticizing Michael Mann for any of that. You'll just seem them criticizing Steve McIntyre for what is a far, far more minor problem. I can't help but read it as:
By the way, after I wrote that post on this topic, I was still quite amused by the things Tom Curtis (and others) were saying. Since people were talking so much about cherry-picking, I decided to finally work on a project I had toyed with earlier. It's a video whose central point is highlighting the fact MBH depended entirely upon cherry-picking a small amount of data. I had written the script a few weeks ago, but I don't have experience making explanatory videos so I kept putting it off.
Anyway, it's still really rough right now, but I could use some feedback:
http://hiizuru.wordpress.com/2014/08/06/feedback-on-a-video/
That it got published in the first place showed the shallowness of the climate clique: The climategate emails prove that none of the paleo community believed in it and only one of the contributing proxies shows a hockey stick (which the original authors said was due to CO2 fertilization) so the result could only be obtained by biased data-mining. But to double-down on stupidity and defend this dross as late as 2012 when the entire climate circus prefers to forget the embarrassing episode and even uber-activist Mann's later work contradicts it, shows a mindset that doesn't care about truth or science - only the message. Maybe if it penetrates their closed minds that this message is leading to immoral policy might they move on. But then again....even that much was obvious long ago.
Anders, employing the Plankton defence yet again!
http://youtu.be/o0_ImQdTrIk
:)
Regards
Mailman
From Anders:
First of all, there is no issue around "how you center the timeseries". If you are constructing PCs you have to center them on a zero mean, period full stop. There is no "other" method that involves some alternate centering. If you do an alternate centering, the result is not a PC, and typically has no known interpretation.
The decentered method can produce hockey stick-shaped PCs out of trendless red noise. There is a demonstration of this in the NAS report around page 90, including a graph of the results. We were told that one of the statisticians on the panel tapped away on his laptop while we spoke, after hearing our assertion on this point, and instantly replicated the effect before we'd even sat down. So there are 2 problems. The faulty method will mine for hockey sticks if even one is present in a matrix of white noise (see our GRL exchange with von Storch and Huybers), or will reorient red noise series to create one even if none of the underlying series has a trend.
You need to read our 2005 summary paper in E&E
http://www.uoguelph.ca/~rmckitri/research/M&M.EE2005.pdf
because it gives an exact decomposition of how the centering issue matters, as well as all the other issues, such as the Gaspe series extrapolation.
One reason it is important to distinguish between pseudo-proxies and pseudo-reconstructions (as Brandon has been emphasizing) is that the reconstruction includes a regression step that dilates the proxy series to fit the temperature record. So it only needs a small blade to work with. The size of the blade of the pseudoproxies will get stretched or shrunk as needed to line up with the temperature series during the calibration step. The comparison of the HSI of MBH98 reconstruction versus the range in MM05 Figure 2 misses this point.
I grant you that the SI ought to have said that the 100 series were picked as candidates that visually demonstrate the phenomenon shown even more vividly and comprehensively in Figure 2. In the race to get the proofs done and the SI submitted I didn't notice that our explanation on that point was inadequate. Having shown in the paper itself the distribution of all 10,000 series, it would be clear to the reader that, had we randomly chosen from among the 5,000 series with positive blades, we'd still end up with a sample with most showing visible blades. And the code was included so there was no intent to conceal the actual selection process. But we didn't say it in words and that is the kind of detail I don't like to leave ambiguous for readers, especially when it gives an opening later on for complaints on non-substantive matters.
But before you or anyone else (Tom Curtis in this case) launches into accusations about "a serious breach of normal standards of publication" you'd better calibrate your outrage megaphone carefully. Show me the volume setting you used when you castigated MBH for failing to report their failed r2 and then denying they'd even computed them when the code clearly showed otherwise, or failing to report the results from the Censored folder and later claiming robustness to exclusion of all dendro indicators, or failing to mention the duplicate use of the Gaspe series and the selective extrapolation in one case but not the other, etc. No doubt the megaphone was set to "whisper quiet" for those, so in the matter of the selection process for the 100 series in our SI, I suggest it should be dialled down somewhat from there.
One final point, for those wanting to really understand the full story, I urge you to buy and read Andrew's book the Hockey Stick Illusion--it's a really excellent read and the story deserves to be remembered in all its fascinating detail. And if you want a quick refresher on the main technical issues, I contributed a chapter about them to the forthcoming IPA volume that you can find on my home page at rossmckitrick.com
This manufactured gotcha regarding the supposed cherry-picking of 100 out of 10,000 series is very tedious. M&M were claiming that the simulations of hockey sticks on trendless persistent series using Mann's transformation "nearly always yielded PC1s with a hockey stick shape". The 100 sticks were simply evidencing that claim. Whether they produced the top 100, the bottom 5, or a statistically random sample of the sticks is immaterial.
And as we all know, once you have a hockey -sticked PC1, then you can claim - as Mann did - that you are capturing the "dominant" climate signal. Do the method properly, and you get a signal that amounts to little more than a footnote.
Thanks for the input, Ross. Very clear, as usual.
lol...I bet the wasps are buzzing on the "Then there's physics site"....what a strange choice of title for a bunch of Jehoveh's Witnesses.
diogenes, you should go over to that blog and read what they have to say. The lack of technical understanding displayed in their discussion of this topic is incredible. My favorite part is how not a single one seems to get the difference between pseudo proxies and pseudo reconstructions. Not a single one of them has indicated any awareness of the fact only three of 22 proxies in the 1400 AD roster were created with "PCA."
In fact, none of them have indicated they are even aware of their being two MM05 papers. Anders apparently wasn't when he first talked about this (even initially refusing to provide a reference to the paper he was discussing), and none of the rest have talked about anything in the 2005 ERL paper. This is telling as even if one felt MBH's "PCA" was appropriate, that would not mean the paper's results were meaningful.
These guys are focusing entirely upon one issue they don't really understand, thinking their knowledge of it somehow allows them to judge a subject which involves multiple other issues.
Anders et al,
A challenge.
Reproduce MBH. Either one.
Using only methods and data available to all.
Publish the steps and results for comment.
Oterwise your comments mean nothing at all.
Morph,
Wow, that's a high bar. I hadn't appreciated that Bishop-Hill required full replication of a paper before anyone could comment. Not sure I've seen it required before. Are you sure you're being consistent?
To be clear, I'm not really defending MBH as such. It's clear that the decentered PCA created a problem. However, it also seems that the results in MM05 (GRL paper) have been somewhat over-interpreted and that some of what they present is not presented as clearly as some might hope (in my opinion at least). Oh, and it now appears the Mann and colleagues had already changed their analysis method by 2005, so quite why MM05 had such impact is somewhat beyond me.
At the end of the day, however, if I want to see our best understanding of temperatures over the last millenium, I'm not going to rely on a paper published 16 years ago, I'm going to look at papers published much more recently. And what do I find if I do that? What I find are figures that appear very like Hockey Sticks. I also discover that a large number of different proxies have been used and that a number of different analysis methods have been used. Does that mean that it's right? Not necessarily, but I'm not going to assume that it's wrong just because there might be some uncertainty.
Oughta read the Bish's book, Anders. You make far too many mistakes. At this rate, you can reach China soon with that hole you're a workin' on.
==========================
Also, Sweetums, it's the shaft, not the blade. You've got a lot of catching up to do before you are persuasive to any but other ignoramuses.
================
Anders, if you are finding figures from recent papers that appear very like hockey sticks, you are either being very selective, or you play a strange game of hockey. If you take a look at figure 5.7 on page 409 of the AR5 WG1 report, you will see a spaghetti diagram that has an evidently marked MWP and LIA relative to the MBH98/99 hockey stick.
Over time, there is a slow but sure retreat from Mann's disastrous work, and I am confident that AR6 will see a further step in the direction of greater variability, particularly since AR5 spaghetti includes Mann 08, which, with its upside down contaminated Tiljander proxies, will surely get dropped next time around (by rights the paper should been withdrawn, let alone get used in the IPCC report).
It’s a shame that the community can’t grow a pair, publicly acknowledge the validity of M&M’s criticisms, and disown the misleading "rebuttals" of Mann and others.
"At the end of the day, however, if I want to see our best understanding of temperatures over the last millenium, I'm not going to rely on a paper published 16 years ago, I'm going to look at papers published much more recently."
They mostly have similar issues. McIntyre has deconstructed a lot of those, as well, and if you read the Bishop's book, you'd have seen that they're covered too.
The main reason the flaws in the original Hockeystick paper are still so significant is not about what it tells us about temperatures, but what it tells us about the process. The point is that the same process that passed the Hockeystick result through peer-review, publication, the climatology community, the IPCC, the IPCC reviewers, and the government's of the world, being promoted to one of the most recognisable icons of the entire campaign, all without a single person checking even basic stuff like whether all the data was correctly labelled and in the right columns, whether the calculation had been done correctly, or whether the putative reconstruction of temperature had any correlation to actual measured temperature at all.
Worse, far, far worse than that, when somebody did finally ask for the data and did start finding flaws, the global scientific community did not leap into action, confirm the findings, scratch it from the record, and detail the measures they were taking so that this could never happen again. No. They refused to acknowledge there was a problem. They defended the paper through thick and thin. They wrote paper after paper attempting to rebut the rebuttal. They refused to release data or cooperate in any way, even threatening to delete data rather than let critics have it, even when it was a legal requirement to release it. They continued to laud the author of the original work. They called the guy who found the problem rude names. And anyone who agreed with him. Even today, the Hockeystick still has its defenders.
(Actually, it was revealed much later that some people in the community had checked the work very early on, and had found the flaws, but kept quiet about it.)
So given that all the later papers came through the same process, (and were in many cases produced by the same group of people by similar methods and using in many cases substantially the same data,) what assurance do we have that any of them are any more trustworthy? What does the long, sorry tale tell you about how thoroughly these results are being checked? And if a result is essentially unchecked, and in many cases uncheckable, why on Earth should anyone believe it?
You see, science only works because all results published are subject to challenge, and only results that have survived a determined and motivated attack get any credibility. The credibility of what survives is founded entirely on the knowledge that if there were any flaws in it they would almost certainly have been found and the paper eliminated from the canon. Thus, for it to be revealed that papers are not being checked, and papers are not being removed even when found to contain errors, it sabotages and subverts the scientific process at its root.
The scientific process is like evolution. Gazelle's are only so fast and agile because lions eat the ones that aren't. Our confidence in scientific results is only so solid and trustworthy because critics in the scientific community eliminate the ones that aren't. But eliminating critics is like putting your animals on a small island with no predators. Your 'science' will soon and with ineluctable inevitability follow the path of the dodo.
igsy,
Sorry, are you seriously claiming that the spaghetti graph in Fig 5.7 of AR5 WGI somehow doesn't look broadly like a Hockey Stick and that the results in MBH98 are entirely inconsistent with that? Are we looking at the same graph.
If this is about the MWA, then maybe you should read what that Chapter actually says about it
Anders
You want a hand with those goalposts?
I thought you said you weren't going to keep on boring us with your views on MBH98 and MM05. Shouldn't have believed it.
MBH98 was a crock of shit, as any half-decent social historian or dendrologist could have told you — and quite a few did.
It wouldn't have needed McIntyre and McKittrick to debunk it if it hadn't been that the climate science community, the eco-warriors and the politicos were so keen on anything that could be used to further their political ends and, provided you didn't look too closely, this fitted the bill perfectly.
Except that it probably turned more doubters into out-and-out sceptics than into believers because it was so at odds with known and well-established reliable historical records and observations and because the lead author soon proved himself to be a still-wet-behind-the-ears newbie PhD. And I still maintain that Bradley and Hughes set him up front and centre to take the flak.
MM05 should have been the final nail in his scientific coffin when they proved his incompetent use of statistics simply undermined his conclusions even further.
And what you are doing now is the equivalent of counting the angels on a pinhead.
I would add my suggestion that you try reading The Hockey Stick Illusion except that I know you won't, if only because as the "science" that underpins the global warming hypotheses gets more threadbare by the day, so the minds of the New Religion's acolytes glue themselves tighter and tighter shut for fear they might learn something that doesn't accord with 25 years of unthinking worship at the shrine of another false god.
Nullius,
The problem I have is that this
is all just a form of conspiracy ideation. We can't trust something because some process didn't work exactly as some people think it should, or some person/people didn't behave as we think they should have. Personally, I have no real interest in such discussions. Until someone can convince me otherwise, they're just a distraction. Maybe you'll turn out to be right and that we'll discover that much of climate science is rife with fraud and dishonest scientific practice. Given the scale and the number of people involved, I really doubt that this will be true and I certainly don't intend to waste my time engaging in a discussion about something that is statistically unlikely (with high significance). You're, of course, free to carry on if you wish.
Here's my final comment on this. You're of course free to use whatever information you like to assess the credibility of a scientific result. At some point in the future (not too distant, I would argue) we will have a much better sense of the credibility of our current understanding of climate science and there is a chance (not small I would suggest) that your view of the credibility will be wrong. I hope that you'll have the honesty to admit that your assessment of the evidence was wrong, and won't be one of those who respond with excuses along the lines of "if only Michael Mann had been nicer to Steve McIntyre", "If only James Hansen hadn't advocated so strongly for action", "if only the IPCC hadn't got the information about Himalayan Glaciers wrong", "if only I could have trusted climate scientists more", "if only Cook et al. hadn't published the consensus project results". Whatever reasons you use to make a decisions about the credibility of climate science today, the decision is still yours to make, and - as far as I'm concerned - you should be willing to own it.
Anders
I'm happy to put it to a vote. Yes, the chart clearly shows a marked MWP and LIA relative to MBH's flat and slightly downtrending line. I hope it is not necessary to reiterate kim's point that it is the shaft, not the blade that is the contentious issue here.
And I'm not quite sure why you're quoting at me that particular excerpt: at least it acknowledges the existence between the MWP and the LIA, whereas the MBH shaft steamrollers them both out of existence.
Finally, more a rhetorical question I suppose, why do so many on your side of the debate implore people to "go read ...", or "maybe you should read ...", etc, when 1) I brought up the reference in the first place, and 2) how can you possibly have the slightest idea whether or not I have already read the text you are telling me to read?
igsy,
I'm not sure why kim brought that up, but okay.
Partly to illustrate that the MWA is not quite as simple as some would to suggest (I have no idea how you see it, though). Also, the MBH shaft only goes back to 1400, so - apart from maybe some blip (non-technical term, I admit) at the beginning, it really seems pretty much the same as the profile in Fig 5.7. Given that the IPCC appears to have defined the MWA as from 950 - 1250, MBH98 doesn't seem to steamroller it out of existence, it seems to not be considering that period at all. Given that the LIA seems to be partly defined in terms of being lower than the MWA, this is presumably not evident in MBH98 simply because their reconstruction doesn't go far enough back to make the actual comparison. There's also the reasonably large confidence interval in MBH98 to bear in mind.
Recap. Mann's work irrelevant by consensus, it's too old and/or artificious. But Mc&Mc are utterly wrong because their paper not 1000% clear.
Later works show HS when based on aforementioned irrelevant paper but their results are very important. And Mc&Mc are utterly wrong because their paper not 1000% clear.
Ipcc consensus drops HS if more than 6 centuries included. This somehow makes recent HS relevant. Also, Mc&Mc are utterly wrong etc etc
we're staring into the anders abyss and it's empty. So is this discussion.
Kim's comment that the blade is not important, is correct---the blade s forced on us by the temperature data. Any red-noise pseudo-data would replicate that. And MBH 99 does go back further, and retains the flat shaft.
More the point if you compare MBH 98 against more recent reconstructions, including Mann's 2008 EIV (his favored reconstruction from that paper), you find that Mann's early reconstruction is inconsistent almost "out of the gate". Remember the horse race starts (and goes backwards along the time access) at the point where the training data begins.
Failure to replicate the training data would yield and unpublishable result.
I think Anders biggest problem is he works backwards from the conclusions he wants to adopt. Which is not how science is supposed to be done. So MBH98/99 is clearly correct, now let's show what McIntyre did wrong, and along the way impute McIntyre's motives for even asking the question is "MBH 98 correct?"
"The problem I have is that this [...] is all just a form of conspiracy ideation."
No it isn't. It's simple observation.
And people misunderstand the issue with conmspiracy theories. They're not a problem logically because they invoke conspiracies (which obviously can and do happen), they're a problem because they present no solid evidence for the belief, claiming that any evidence has been "covered up" by the conspiracy itself. It's the lack of evidence that is the problem, not the conspiracy.
In this case, we do have solid evidence.
--
The paper is crap.
It went through all the official checks and the fact it was crap was not detected.
Therefore the official checks do not distinguish crap from true results.
--
A sceptic identified it as crap and said so publicly.
The climate science community did not withdraw their support for the paper, even after this was published.
Therefore the climate science community are *knowingly* supporting crap as science, instead of eliminating it.
--
Our confidence in science is founded on the belief that scientists are thoroughly checking for and eliminating bad results as a matter of principle.
They're not.
Therefore confidence in climate science is unfounded.
--
I don't have to posit any secret conspiracy for evidence of this - I can rely on a scientist's public statements. If he knows all of the above and doesn't condemn it, I can be absolutely certain he stands in opposition to basic scientific principles and standards. If he makes statements in support of climate science against its critics without first finding out whether those criticisms are valid, I can be absolutely certain that he does not follow basic scientific standards. He's just said it himself.
And if thousands of scientists do so, then thousands of scientists are condemned - by the evidence of their own words. There's nothing secret or hidden about it. It's right there out in the open.
What is *not* so clear is why they did it. Personally, I don't think a conspiracy is required to explain it, just human nature. People get sloppy about standards if they're not held to them. They make assumptions. They follow the herd. They all assume, as you do, that the rest of the herd can't all be wrong, and they all assume that the rest of the herd are not all doing what they're doing and that someone, somewhere, must have checked.
It's called groupthink, and all humans are subject to it, scientists included. That's why we have such strict standards in science about not trusting any work until we are sure it's been thoroughly checked, and not taking anybody else's word for it that it has.
The idea that this is statistically unlikely is based on the assumption that each scientist comes to their opinion independently. But if each scientist, as you just did, derives their opinion from what all the other scientists are saying, then the independence assumption is not valid, and the probability is much higher. And again, I'm not speculating without evidence here. You, and every other scientist who follows the consensus, always gives the same argument: "thousands of scientists can't be wrong".
And what if they're all just like you, and saying the same thing?
"I hope that you'll have the honesty to admit that your assessment of the evidence was wrong, and won't be one of those who respond with excuses..."
Even if it turns out that I'm wrong about the conclusion, I'm still right about the evidence.
If someone is asked to reduce the fraction 26/65 and says "we cancel the 6 on the top and bottom of the fraction to get 2/5", they're wrong. They're still wrong even though they got the right answer, because their method was wrong, and right or not, results obtained by a wrong method are unreliable. Science does not require that one merely be right, you have to know that you're right - your confidence must be justifiable. There's no credit for the conjecture, only the proof.
So if it turns out that the world was in danger after all, and we didn't know until too late because of these clowns fooling around hiding their data and sticking their thermometers next to aircon units, I'm going to be angry. We told them to fix the science ten years ago. We even told them a lot of what they needed to do. But instead, they all preferred to pretend there was nothing wrong, and stonewall the scientific criticism. They put their own careers ahead of the planet.
Personally, I'm not too worried. The success record on 'End Of The World' claims is not good, and nor is that on visibly bad science. Nor do I think that they themselves are too serious about it either - they still lead high-energy lives jetting off to climate conferences and stuff. If government's were serious about it, they'd have just gone nuclear and have done with it. And the international negotiations are stalled not because the rich industrialists are insisting that that the world should be allowed to continue emitting CO2, but bizarrely because the climate activists are. The Byrd-Hagel resolution identified the real issue back in 97 and it's still true today.
But I'll stand by my principles, and what I said, however it turns out.
Anders
I appreciate your reasonable tone in response to me, but I must qualify my appreciation by saying it is disappointing you felt the need to play the conspiracy card in your reply to NiV's excellent post.
You might note that I referred to MBH; I guess that there's a kind of unstated convention that when we talk of MBH, we really mean MBH98/99. The 99 paper extended the 98 paper - with basically the same techniques etc - back to the year 1000AD, so it self-evidently does cover the MWP.
In citing the IPCC report I was erring on the conservative side with respect to the variability question. Looking at individual reconstructions, we see that Christiansen & Ljungqvist, Moburg, Loehle & McCulloch (yes, it is included by the IPCC) really do by any standards have a pronounced MWP, even if in others it is not so marked.
I am not saying all recent papers have a MWP (or MCA if you wish), let alone "prove" there was a MWP/MCA. However, some influential papers clearly do - and if you examine them, you find the C&L paper, for example, has a spatial diversity that is lacking in many others, not to mention the fact that the bristlecones do not exert an undue influence over the shaft.
Finally, there is this excellent blog post that I cannot help but recommend :-)
http://bishophill.squarespace.com/blog/2012/10/27/a-warm-welcome-back-to-the-mwp.html
Omnologos,
You can think that if you want. I certainly don't. Who says 1000% anyway?
Carrick,
You can think whatever you like about me. You're of course, not guaranteed to be right. I think you're annoyed that I wouldn't agree with you about Cook et al.being unethical and are now just going to go around bad mouthing me because you're annoyed that I didn't bow down to your self-declared superior intellect and absolute sense of moral superiority. I may, of course, also be wrong. I have seen nothing to convince me that I am though. Where does your dropbox reconstruction come from anyway? Doesn't appear that consistent with the recent IPCC report. Also, have you ever heard of confidence intervals?
igsy,
I did it because that's what it sounds like. Nullius may well be correct, but if someone is going to claim some kind of conspiracy, at least have the courage to own it. Implying something with respect to a field in general just because a particular process wasn't perfect or because someone's behaviour wasn't as ideal as we might like it to have been is - in my view - a form of conspiracy ideation. It's a suggestion that collectively a group are allowing unsatisfactory behaviour to negatively influence their work in general.
Just to be clear, I'm well aware that there may be more variability in the last millenium than may be present in some reconstructions. I'm also well aware that such understanding such variability can play an important role in understanding the role of different factors (volcanoes, solar, internal variability) in climate variability.
As you yourself appear to be acknowledging, even now there is a different level of variability in different reconstructions. Making a big deal out of an early version that did have some issues with the analysis just seems like a distraction and rather irrelevant.
"It's a suggestion that collectively a group are allowing unsatisfactory behaviour to negatively influence their work in general."
It's not a suggestion, it's a statement. The original Hockeystick paper is junk. It's a mathematical fact. And yet there is a group of scientists who have gone to a lot of effort over the years defending it. I'm not going to try to make any assertions about why they have, but it's clear that they have.
You know they have, too. But you don't want to accept it because you don't like the implications. You think there must be something wrong with the argument - because it contradicts the ad populam argument - but you don't know exactly what. How could such a lot of scientists be wrong? It's impossible! So you would rather reject the mathematical arguments showing that they are.
Virtually *all* the constructions are flawed. The reconstruction methods all work by searching the input data for series that correlate (or anti-correlate) with the modern temperature rise, and concluding that those that do so are more accurate temperature records, while those that don't aren't. The ones that do are therefore weighted more heavily, flipping them upside down if necessary, and the ones that don't are multiplied by numbers close to zero. The result is that all the weight is put on those few series with spikes at the end, which are often there for entirely unrelated reasons. We know they're not responding to temperature, because the temperature records where the tree actually is are not rising during the modern period. We know they're not responding to temperature, because when you compare the reconstruction against measured temperatures during an interval not used for training, there's no correlation. None!
The MBH series used a lot of input series, and then used the biased PCA process to mine for the few end spikes. Most of the other reconstructions achieve the same result by using fewer series, and carefully selecting them to assure that there are a few spiked series in there. There aren't all that many of them, so they turn up over and over again in reconstructions. Gaspe, the Greybill bristlecones, Tornetrask, Polar Urals, ... over and over again. And they don't correlate with temperatures outside the calibration period, either. Rather famously, a number of them even diverge rather dramatically in the modern era, when we know what the temperature was doing. They appear to say that after 1960 the temperature was going down. The climate scientists' response to this blatant failure of their methods to measure what they're claiming to measure? Cut the offending data off! Truncate the divergence to hide the decline!
It's unbelievable! And *that*, I think, is the problem. Nobody can believe they could have got away with it if that was what they had really done, so everyone assumes they didn't do it. There must be some good reason. And because everybody else assumes that too, and therefore doesn't say anything, nobody can believe either that nobody else would say anything. (Sceptics don't count.) So they don't.
If the problems in the "early analysis" had been acknowledged and the reconstruction struck from the record, as *should* have been done, you'd be right. There would be a bit of a fuss over how it had got so far, but it would have been over and science would have come out ahead. (Even better, if the problems had been picked up in review, the graph would never have become as famous and as vulnerable as it did.) But the problem is that it didn't happen. Scientists did, and still do, defend the errors. Anyone can make a mistake - it's refusing to correct it that causes the trouble.
Anders, the issues with Cook have to do with ethics of science issues. You seem to be gobsmacked that anybody would even think that is important. And of course, as is your wont, you impute the motives of people who ask questions that you don' see as important. Or they are ideating conspiracy. (Because in Anders universe just being Anders is enough to be an expert at psychology. "The Doctor is In.").
And that all makes me arrogant, eh?
I never assume I'm the smartest person in the room, in fact I know I'm not (though I'm lucky to work with very bright people), but I do ask people to substantiate what they claim and don't take them very seriously when they either don't or can't or especially when they have tizzy fits when they can't.
And I do expect them, if they claim to be practicing scientists, to understand the philosophy of science well enough to understand why ethics in human subjects research is important and ethics codes on responsible conduct in research are issues that should be explored.
Regarding the figure… of course I generated it. It's not difficult to do. You download the various time series. You resample them so that you can combine them as an ensemble and compute confidence intervals. It's ten minutes of work, doesn't require a Ph.D. to produce, or peer review to confirm its validity.
Scaling bias is an issue that you have to address before you can combine them. What I did was simple regression to put them on a common scale, total least squares regression would have been better. I aligned them using a common baseline (Loehle is an issue there because it doesn't extend to the Common Era.). That doesn't affect the averaging process, but does affect the estimation of variance.
Regarding confidence bias… gee there's nothing arrogant about the way you asked that question, is there? (I wonder if you even realize what a putz you are.)
Yes, of course I did a study on that too.
Here's a figure (1-sigmas). Blue is the ensemble, green is MBH98.
It doesn't take a rocket surgeon to figure out these two curves do not show the same thing. But in case it isn't obvious,
same figure shown as z-scores.
Anyway, the real problem with MBH98 is a loss of low frequency information:
It ends up being filtered red noise.
It is not a valid reconstruction.
Nullius,
I disagree, there are no "facts" in this context. There analysis method had a flaw. It's no longer used. Drawing much broader conclusions from that is your right. I think it's a distraction.
How do you know what implications I'd like? It would be much easier to accept that these temperature reconstructions are flawed than accept what they're currently illustrating. In what way do I benefit by accepting this evidence?
Anyway, I'm gathering that this conversation is going the direction of most conversations on this topic.
Pin dropping is never unexpected. Anders has no interest in the science of Mann or Mc&Mc...there are too many errors in his description of it, no attempt to understand the Bish's book, and his hand waving on the mwp would've marked him as skeptic until 2y ago. He just doesn't know the topic and shows no inclination to learn about it.
Only remaining reasons to opine about how wrong it was the response to a now irrelevant paper? Character assassination of Mc&Mc springs to mind. Likewise being able to proclaim Mann a visionary, and resurrecting the hs in order to revive the comatose push for climate change mitigation.
I see no conspiracy in any of that. Like good fascists in Italy and good communists in the USSR, True Believers know what they have to think and do, without prompting each other. Only necessary condition is a stomach for hypocrisy for a higher purpose