I might as well illustrate real-time frustrations in dealing with the Hockey Team. I’ve written back on Jacoby’s stonewalling to Climatic Change (who have been trying and who at least got a refusal from Jacoby) as follows.
After receipt of the Jacoby response posted up yesterday, Climatic Change asked me to deal directly with Jacoby, saying that they had done all that they could do. Obviously, if Jacoby had been responsive in the first place, I wouldn’t have involved Climatic Change (who published the articles in question). Here’s my reply to Climatic Change.
Dear Dr. Schneider and Ms Kivel,
I contacted Jacoby directly without any acknowledgement, which is why I sought the intervention of Climatic Change, who published the Jacoby material. While you may feel frustrated in this matter, I have been seeking this information for over a year and still have gotten nowhere. I wanted to arrange for re-sampling last summer and will try again this summer.At this point, we know that Jacoby attempted to re-sample the Ste Anne River, Gaspé site in 1991 and did in fact re-sample cedars in Gaspé in 1991, presumably close to the original sampling location. Let’s go through this in baby steps. Can you request Jacoby to do the following:
1. Provide the directions and maps that Jacoby used in 1991 in attempting to locate the Ste Anne River site.
2. Provide the directions and maps that Jacoby made in 1991 to document the location of 1991 sampling.
3. Provide the measurements made in 1991.I apologize for continuing to involve you in this, but the fault lies with the original authors. As the journal of record, you still have responsibility in this matter. You have the possible recourse of retracting the Jacoby articles if he will not provide location information. If nothing else inspires cooperation from Jacoby, perhaps it is time to remind him of this possibility.
Regards, Steve McIntyre
Climate Audit 
78 Comments
Any bets on whether Steve gets a useful reply?
$10 says stonewall.
John, This is an important request, clearly stated, pertinent and reasonable. Even proper. You think toss up odds?
I can see it now… I’m sorry but the paper data was burned up in a fire caused by an overheating AC unit. Our lawyers have directed us not to release any electronic version of the data either pending their decision on whether or not to sue various energy corporations for releasing unwarrented quantities of CO2 into the atmosphere, thus causing the Anthropic Global Warming which produced this overheating and thus the loss of scientific data.
Regards, the Hockey Team.
Re: #2
John, I have a 20-sided die. Don’t make me roll initiative…
John, at $10 it’s a sucker’s bet so no thanks. It’s such a sure thing that it’ll look more like a donation than a bet. The hockey team might get wind of it and claim this site is funded by gambling activities.
This makes me think again and again of the Bellesiles affair in a very different field: US personal firearms history. In the end, Bellesiles was discredited by his own university and his own profession. If Mann et al. or Jacoby et al. ultimately fail to provide code or data, I hope the profession reacts accordingly.
Re: #5
You say that like it’s a bad thing. It would be an improvement upon secret fossil fuel industry slush funds. (And in case there are secret fossil fuel industry slush fund donors out there, where the hell are you?)
When I was an astronomy major at UT-Austin back in 1977, I was invited to hear a graduate student defend a claim to have located a good portion of the universe’s missing mass. You see, our models said it was out there, but where? He thought he could detect it in halos surrounding galaxies. In the end his claim didn’t hold up. The exchange was as fair as anything I’ve ever seen. Had a hint of bad faith on his part crept in, blood would have been on the deck. Not so in the climate science club, unless they have redefined bad faith.
Steve: on 4 April (www.climateaudit.org/index.php?p=66#comments; posting 88),
you said:
> In the future, I will adopt the stated realclimate policy of limiting
> posted comments to questions about science and technical matters.
and then censored my following posting which directly addressed the thread in question.
So ….. tell me, Steve …… do you really believe that the pathetic, childish and cackling drivel of the present thread counts as "science and technical matters"?
Or is it one rule for the "friendlies" and another for those who dare to criticise you?
Steve: The replication and availablity of data is something that I think is very important and has been a theme right from the beginning.
John H, #2 was a serious question. In #8 I meant to say that any sort of deliberate foolery to hinder skeptics would limit or end a career in astronomy back when I was exposed to it. It can’t be science without replication to start.
John Hunter, Are you trying to say that “the pathetic, childish and cackling drivel of the present thread” would only count as “science and technical matters” if the amount of money being wagered on our probability analysis was more than $10? How about billions of $$$? Would that make it “science” as defined by your circle of friends?
Steve and John allow a LOT of leaway in the content of these comments as witnessed by the postings of Peter H. and for that matter yours. This is a very open and honest discussion of the facts and issues so you must have gone WAY OVER the line in terms of a personal attack for Steve to have censored a comment. If there was any factual merrit in that comment I invite you to censor out the diatribe and post a sane and sober version in this thread so we can see what facts were suppressed. Your next posting on this site should be very interesting.
A note to Steve and John. As an interested participant on this site, please ensure John Hunter’s next comment is posted as is. For that matter, if it is still available, I think many of us would appreciate seeing the original “censored” comment so we may gain insight into John Hunter’s thought process.
Steve: I asked you two questions in posting 9. In a tradition which is all too common on this site, you have answered neither.
Steve: : 1) The availability of data is a technical matter. 2) As I recall, I excluded a post of yours going on about John A.’s personal information. He does not have independent means and is worried about repercussions from people like you. The topic has been discussed and I’m not going to debate this endlessly. If John A changes his mind, I’ll let you know.
I presume that you are unable to criticize any scientific comments made here or else you would have.
I am not sure what you are all implying, conspiracy or lack of rigor, but let me tell you a little story about science here in the UK. As you may know we have had a small problem with a disease known colloquially as Mad Cow disease. There was a lot of concern that this may have spread to sheep, but had not been detected because of a similar disease called scrapie. The UK government gave a grant of about £2 million for a prestigious institute to investigate. Only just as they were about to publish their results did someone notice that they had been looking at the wrong set of brains (turned out to be a labelling mix up in a freezer). This caused much embarrassment to all concerned and lots of questions in newspapers. First point, everyone makes mistakes, even when their work is very important, data can easily be lost. I myself have lost data which has been archived on floppy disk because the disks decay. Record keeping is not always what it should be and when people leave they throw away things they don’t think are important, or leave things which no one else can understand. Second point, as a result of the Mad cow debacle all research institutes etc. doing UK government funded research have to comply with auditing procedures. Where I work every piece of lab or field work has to be entered in specially printed notebooks in a certain way, all samples must be cross referenced with computer files and notebooks. Every piece of equipment is labelled and serviced as part of QA. These sorts of procedure should be common practice now in climate research as well, I would have thought, but I doubt they were in place 15, 20 or 30 years ago.
# 11
Michael, you must have remarkable powers, or very good contacts, if you know what John Hunter posted…
As to my comments, feel free to point out where I’ve gone over the top rather than responded in kind. Of course you are one of the several here who seems to reply to me by just casting aspersions in my direction 😦
Re #12
Paul,
You’ve made Steve’s point. If data is used to support public policy then it must be properly audited and it must be archived safely. This is no more and no less that Steve’s entire thesis on this weblog.
Re #14
Except you need to explain how you can impose things retrospectively? By time travel?
Paul is right, the implication of most contribution here, indeed of this site I think, is that there is a conspiracy at the heart of climate science. At the very best such accusations are a massive exaggeration of some minor cock ups. And, again as Paul Gosling point out, the major mistake came out re CJD. It would with climate science, indeed I think it would *have*, if it were there.
Peter, I was refering to your many examples of “failing to engage in the facts”. Your ability to misinterpret even the simplest of statements is becoming legendary on this site. Most of your comments have been nothing but personal attacks. Despite the level of vitriol you have demonstrated in some of your comments I have not seen any indication of you being censored. That gives us all a good idea of how much Steve will tolerate. If we assume a reasonable amount of guard banding then Hunter’s comments must have been very extreme to warrant censorship. See? All it takes to deduce the nature of the “censored comment” is the ability to read and comprehend the English language and to have at least a modest reasoning ability. It doesn’t require psychic powers or being part of a secret conspiracy.
I don’t get it.
My (and my peers’) privately-funded master’s and doctorate research had to have its guts spilled-out all over the place to make sure every “i” was dotted and every “t” crossed. Any methods I or my peers used had to be described in gory detail and be completely acceptable (I witnessed a peer’s research go down-in-flames when his calibration procedure was brought into question, even though it was widely-used in his field). Every single data point used had to be available in one of the appendices. Any new methods developed were either the intellectual property of the university or the funding agency – not my own or my professors’. And for publication, even though things were a bit more streamlined, papers needed to have plenty of supporting evidence. Diskettes were sent containing all of this data to the journal along with paper submission. Duplication of results was important enough. The capability of other researchers to replicate the results was even more important.
But when it comes to publicly-funded research being done at universities and used to determine extensive political and economic policy across the globe, concealment is allowable? Methods are the property of those doing the work? A lack of replication is permissible?
I don’t quite understand why my research – which in the grand scheme of things was quite a negligible contribution to the scientific world – was held to a higher standard than research which determines global policies.
Clearly your experience of graduate study was very different to mine, when no one gave a damn what you were doing as long as you finished on time. Though you don’t say, I suspect that you were working as cheap labour, doing R & D for a (chemical, biotech?) company. If they were investing their money in your work they wanted to know it was correct. Most scientific research (including what I do) just keeps us off the streets. I think you are naive to think that government policy is decided on the basis of science, its based on what will get the most votes.
Paul,
The problem is therefore what the science is, that informs those votes, which demands the policy.
I’m interested in some of the comments in #12.
I tend to agree with the general philosophy that “mistakes happen” in science – papers submitted to journal are subject to pressures of time and resources and aren’t always perfect, and the review process doesn’t always pick this up.
But I’m not so convinced about the fact that the control standards applied to papers were less rigourous some years ago. OK I can’t comment on papers from 30 years back (I was more interested in filling nappies 30 years ago than in scientific analysis) but certainly 20 years back the importance of good configuration control was well understood, particularly in the commercial world. Tight configuration control can be tracked back in the organisation I work for, and I can recover detailed info from analyses dating back into the 60’s and 70’s through the configuration management systems we have in place.
Perhaps though one of the differences is again coming from a statistics viewpoint – I know you aren’t a huge fan of statisticians, but any scientific study involving stats MUST have a far higher degree of rigour applied than in other areas of science. This is well known (and has been for many years) for the primary reason of eliminating risks related to subconcious data selection.
It may well be that recently climate science has ventured more into statistical methods more recently and that the necessary minimum standards of this field have not come across with it. But this creates a serious problem in the trust in the data. If a data set cannot be replicated or traced, including both accepted AND rejected data sets, it should not be viewed as an unbiased estimate. Hence my discussion on Study Protocols in comment #12 of “Jacoby’s “Lost” Gaspé Cedars”.
I also recognise that effectiveness of configuration control does vary between scientists and often certain individuals are better than others. It makes me suspicious that almost all of the hockey team seem to lean on one side of that particular fence.
Last comment – with respect to mad cow disease, do you remember the media headlines that nvCJD was going to become endemic in the UK, how we were just on the leading edge of a bell curve, with wild projections of death tolls in years to come due to long incubation periods etc. etc. I can’t help but make comparisons with the climate change lobby. Those predictions never did come true, and the total level of all CJD deaths remains at a similar level to which it always did. nvCJD was never formally linked to BSE, although there was some circumstantial evidence to support the hypothesis, there are now stories of possible nvCJD cases dating back to the 1970s. These stories (“oh actually we were wrong, there is no crisis”) never did make the headlines as well as the “we’re all doomed” stories did. My suspicion is that climate change will follow a similar path.
Paul,
I don’t know where you’re at, but here in the U.S., all graduate school labor is cheap labor! In some cases, it borders on slave labor 🙂 It was bioremediation studies that I was involved in – feasibility, rates of degradation, identifying metabolites/end-products, etc. And the people funding the project weren’t the ones concerned with being thorough – it was my professor/research advisor, my thesis committee, my department, and the university (and, of course, peer-reviews from journals for publication and organizations for presentations). The funding people were more interested in graphs and presentations about the research. They were essentially lay-people who didn’t have the first clue about the importance of things like duplication, replication, statistical significance, etc.
I am not naive enough to think science dictates public policy. I happen to find this to be a tremendous flaw.
Oops more mistakes on my part!
In comment #21, I start by saying I’m responding to comment #12, when in fact I am (hopefully obviously…) responding to Paul’s comment #13. If only I could get my comments peer reviewed(tm) before posting they would clearly be “climate change” perfect. 😉
#21 Spence
I’ve had doubt about mad cow disease for some time. Clear something *did* infect (if a CJD can infect?) the UK dairy herd. It was a very nasty CJD, almost certainly caused by feeding ruminats rendered sheep in the 80’s, I seem to remember controls on such practices were relaxed? Did this nasty CJD pass to people? I think the jury is more out than it was over whether a lot of people are susceptable to it.
That said I’m as convinced about the rightness of ghg thoery, of our measurement of CO2/temperature increase, and of the likely outcome this century as I have ever been.
You most now decide whether to describe me as open or closed minded. Best of luck 🙂
#22 Paul, Science is amoral. Wouldn’t it be better if science informed, not dictated public policy? No one here wants eugenics. Being a slave to science would be hell. Dictates was too strong a word right? Otherwise I’m with you.
Steve: as regards posting #12, you once again respond to my (quite reasonable) previous requests for “disclosure” of the identity of one of your workers with insulting innuendo. You say that “he does not have independent means and is worried about repercussions from people like you”. If you have any evidence that I have used or would use such information in an underhand way then please provide it — otherwise please retract your suggestion.
You also say “I presume that you are unable to criticize any scientific comments made here or else you would have.” Your memory may be poor, but you should at least be able to recall a reasonably protracted discussion we had concerning your confusion of PCs with actual reconstructions. In this regard, perhaps you would like to comment on the following statement from the “2005 Index of Leading Environmental Indicators” at http://www.pacificresearch.org, which says:
“Scientists have shown that the process used to generate Mann’s graph would generate the same result from any series of random numbers.”
I take this to mean that, if I gave Mann a set of proxies constructed from random numbers, then the resulting RECONSTRUCTION would ALWAYS be a hockeystick.
Do you agree with this? If not, are you going to address it, audit it and have it corrected?
I don’t really want to go on about this as its rather off topic, but there is still a big question mark over vCJD in the UK. All those who have died so far (149) have one version of a relevant gene the other two thirds of the population have the other. They may be immune or highly susceptible but have a longer incubation period.
As for the scientific rigour arguments. In my experience most experiments could have been done better with hindsight, but we don’t have the time or resources to repeat them, so we publish what we have knowing it to be imperfect, because if we don’t then we are likely to loose our jobs (much more so in the USA I understand). In most cases this does not matter because if we have published something stupid it will not be replicated and it will fade from the literature. Perhaps climate science should be treated differently in view of the consequences of getting it wrong. Trans-national groups doing the research with (almost) unlimited resources, which include sceptics such as Steve. But this is probably wishful thinking. As an aside, the review process — how long do you spend reviewing a paper, has this got shorter over the years as work pressure has increased? My boss allows an hour, I try to spend a day (I am not as busy as he is), is this enough, should we be paid to do it and be more rigorous?
Humm, I’ve just read #26 and then re read #12. This bit of #12 is relevant (and personally directed): “He does not have independent means and is worried about repercussions from people like you.” Repercussions? People like you? What *DO* you mean?
Whatever, the implication is clear. While you and John A have characters as pure and perfect as the driven snow, John Hunter is ‘people like you’ and a worry for John A.
I don’t think any of us know *anything* about John A – I certainly don’t. Yet, you, Steve know enough about John Hunter (somehow) to ALLEGE he might be a worry for John A. How??? I think that is, especially for a scientist about a fellow scientist, a ungracious thing to imply or allege 😦
I must say, since I’ve been as open and honest about myself as anyone here (and many people here could identify me), that *I* feel rather worried by your comment. What repercussions might I face from people like you and John A as a consequence of my words here?
Can you, and John A., assure me that people like you are not a worry for me? I suspect not…
Copy taken for reference, since I have doubts you’ll publish.
RE#26: “I take this to mean that, if I gave Mann a set of proxies constructed from random numbers, then the resulting RECONSTRUCTION would ALWAYS be a hockeystick.”
See http://www.climatechangeissues.com/files/PDF/conf05mckitrick.pdf, page 10. Does a “pronounced hockey stick-shaped PC1 over 99% of the time” in 10,000 repetitions satisfy your “always be a hockey stick” criteria?
Michael, I’ll be honest, I haven’t seen where the owners of this site have invited JH to re post. Where is it?
Children, children… let’s not flame.
Peter, it’s not a matter of trust, it’s a matter of replication. I’m surprised that it’s an issue. Any corporation offering securities to the public has to be audited. As I often mention, I sometimes feel like an anthropologist in the land of academics, since my experience is different. For example, a one-hour peer review is a lot different than the audit of a bank. People often point out that business audits don’t always work e.g. Enron. But that doesn’t mean that you shouldn’t audit financial statements. My point is that “peer review” in climate science is a very limited form of due diligence and that, prior to applying “unaudited” results in public policy, some form of engineering-equivalent due diligence is surely needed. It’s not a matter of trust.
Regards, Steve
Re #26: John H, thanks for the reference to http://www.pacificresearch.org
Their publication “2005 Index of Leading Environmental Indicators” is a positively welcome relief from the outpourings of government-funded doom and gloom agencies.
Re: #27, sorry about picking up the vCJD discussion, I think it is an interesting case study in respect of both how science conducts itself and how media and politics respond to science, but you are right in saying it is “off-topic” for this discussion.
I recognise a lot of what you say regarding reviewing papers, although I would stress the difference in rigour required for statistical studies over “normal” scientific studies. The reasons for this are simple: statistical studies, if not properly conducted, are always open to dangers such as subconcious data selection. For this reason, a different level of rigour is required, and usually applied. This is why study protocols are wise, together with archiving all data (both accepted and rejected data). Of course even this is open to abuse but it prevents the subconcious introduction of errors or bias. This is important, as such things can quickly and easily invalidate basic statistical tests, which carry an in-built assumption about sampling, enabling false results to be reported.
Peter, I disagree with this. Strictly speaking, through the anonymity of the internet, we don’t really know anything about people who post to blogs. So it is not a question that Steve knows enough about John H to assess him as a risk, more that Steve knows insufficient about John H to assess him as being safe. The two are very different comparisons (as any good statistician will tell you ;))
I can fully understand the desire to remain anonymous on these topics. I would say the environmentalist lobby are second only to the animal rights lobby in the UK for extreme (and often illegal) tactics to impose their viewpoint on others. Being from the UK you may be aware of the recent scandal in which animal rights activists stole the remains of the dead mother of a farmer who bred animals for scientific research. Now I’m not saying anyone is here to do that, but with these kinds of sick, extreme activities becoming more common in the UK, coupled with the extreme lengths environmental protesters could be willing to extend to, I fully understand John A’s desire for anonymity.
I stress (although I shouldn’t have to) that I am not saying John Hunter (or whoever is posting as John Hunter) is likely to do this, but just pointing out that through the anonymity of the internet, we cannot really be sure of who we are communicating with, and a request for anonymity is not so unreasonable.
If technical discussions are required, Steve is available and easily contacted. Otherwise I think it would be better if we focus on the science, not on the personalities involved.
Re 30, 31 & 32
Peter, in #11 I asked JH to repost and I asked Steve and John to let it through so we could all see what JH was so upset about. There has been no objection from Steve or John but JH appears to have declined the invitation in favour of more of the same old same old. Oh well…
As for the colloquial definitions of BSc, MSc and PhD, they were taught to me many years ago by a PhD who had earned my respect. It is most unfortunate that he has never needed to change his opinion of credentials. He earned his so he could get his foot in the door of his chosen field. It is the only time, that I am aware of, that he ever used them.
There is nothing inherently wrong with a PhD and I do know many fine PhDs. It’s just that too many people who earn them never live up to them and it tarnishes the reputation. Maybe if they needed to be reearned every 10 years…
Re #35
I have to admit that my research career has been almost exclusively based on nice replicated experimental work, requiring fairly basic statistical analysis,
rather than complex multivariate analyses, which is why I tend to stay out of discussions about statistics. However, though Mann et al 98 has been discredited, the rest of the hockey team have produced similar results. Are you saying that they are all making the same statistical errors, or is the argument now about the validity of their data sets? The big problems seem to be with the bristlecone pines and the Gaspe cedars, do all the hockey stick reconstructions rely on one or both of these data sets?
Re #38:
Paul, this seems to be the next order of business. In my opinion, none of these other studies is robust at all, although the details tend to be different. As an aside, the other multiproxy studies are not "independent" as ordinary people understand the terms, either in authorship or proxy selection. If you look down the masthead, you see that Jones, Briffa, Mann, Bradley, Hughes, Cook are coauthors in pretty much every multiproxy study. If you look at the proxies, they overlap.
I’ve got a long of work in hand in Jones et al 1998. You can probably see where I’m going with this from my postings on Polar Urals and Tornetrask. Jones et al 1998 has only 3 proxies in the 11th century – 2 of which are Polar Urals and Tornetrask – each of these subsidiary studies has lots of problems. In fact, if you take out the section of the Polar Urals record based on only 3 badly crossdated cores, you get a quite different result. Polar Urals is used in nearly every study.
Jacoby’s northern treeline study is problematic: if you simulate the choice of the 10 "most temperature sensitive" of 36 sites by simply picking the most hockey stick shaped series from 36 red noise series, you get something that looks like Jacoby’s reconstruction. This reconstruction is used endemically.
Briffa et al [2001] shows declining 20th century values after 1950 and simply invents a hypothesis of an unknown anthropogenic impact. I’m going to post up on this. They usually delete post-1960 values in spaghetti graphs with this one in. The 387 sites used in this study are drawn presumably from Schweingruber collections but I have been unable to get a listing of these sites to analyze further.
Briffa also notes a 20th century decline in ring widths and densities over hundreds of chronologies. Yet when you look at the tree ring sites selected into multiproxy studies, you will see sites with big 20th century pulses. Yamal has a late pulse and gets included. There sure seems to be cherry-picking, but it’s hard to quantify exactly. This is something that I’m working on.
Thompson’s Dunde data has a hockey stick pattern and recurs. There are different grey versions and, 17 years after drilling, the sample data has not been archived other than decadal averages archived last year in response to my complaining to Climatic Change. There are problems with this series.
Crowley and Lowery 2000 uses 2 bristlecone pine series, Polar Urals, Tornetrask and Dunde. These impart the hockey stick shape to CL2000. CL2000 actually has more of a Moberg shape with a MWP similar to mid-20th century.
Mann and Jones 2003 use the North American PC1 (bristlecone pines). I don’t know how the weighting works in this, but my surmise is that the hockey stick shape is an imprint from this.
The data from Esper et al 2002 is not archived. Here I’m interested in the impact of the Modern Sample Bias, noted by Melvin and which I posted on.
I’m also interested in altitude changes in tree ring sites. It makes no sense to me to ignore altitude changes, as Briffa does at Polar Urals. This also seems to affect bristlecones. How can one possibly reconcile the evidence of high medieval treelines (well above modern levels) at both Polar Urals and bristlecone sites with claims of coldness in these periods? Especially without a systematic discussion.
In my opinion, there’s a lot of hair on every one of these other studies.
Regards, Steve McIntyre
Re #38
Steve has already given a very thorough discussion on this topic, but I would try to summarise my views on this:
1. These discussions are often about the input data to these studies, which is often common to several of the reconstructions; so statistical flaws may affect more than one study.
2. The historical reconstructions really don’t agree well at all – within the bounds that the proxies measure temperature over (which may be limited – see 3.) the spread of the reconstructions is as large as the underlying signal
3. We make the assumption that the proxy response is linear and can be extrapolated – this may not be the case. Most of the reconstructions stop around 1980, and then only have the CRU temperature beyond this. But the proxies may not register higher temperatures in the same way the instrumentation does. This is why it is critical that the proxies are bought up to date – if the current “highest” temperatures are not reflected in the proxies today, why should it show up 1000 years ago?
4. The von Storch paper indicates that multi proxy studies may underestimate past temperature fluctuations, another reason to question whether the conclusions being drawn from these reconstructions are valid or useful.
I would not expect the dendrochronology data to have a linear relationship to temperature at either end of the temperature extreme. At the upper end drought is likely and at the lower end i would expect late frosts to have a disproportionate effect. Though this is only supposition, not from direct knowledge.
Re#41: “At the upper end drought is likely…”
Is it any more likely than rainy deluge? I thought a warmer world was supposed to be a wetter one.
Paul, I agree that significantly cold temperatures could lead to a pproblem with frosts and other inhibitors of growth, however i would be more cautious on the immediate linking of heat with drought. In most of the sites under question, the overall high temperatures are not of the sahara desert level, and increases in temeprature could well be associated with additional precipitation rather than a rediction. However the idea that temperature is not the primary driver is indeed worth serious investigation, as it seems quite evident that at some sites such as the Bristlecone Pines, that is indeed the case.
What does seem (and is still being followed up as Steve indicates) is the feeling that most reconstructions seem to assume that any modern spikes in the dendro record are temperature related, while any “non-spikes” are due to other factors preventing the temperature spike showing through. The records often seem selected on this basis, Jacoby being a prime example.
So, almost regardless of what the results are, this research is worth doing. Poor science should not be covered up because it provides the resulots some people want to hear. If the reconstructions are solid, then a properly carried out investigation will show this. Replication is a valuable technique in contentious areas, and getting down to the data is essential.
Once again Steve ignores the difficult questions. I asked in posting 26:
“I take this to mean that, if I gave Mann a set of proxies constructed from random numbers, then the resulting RECONSTRUCTION would ALWAYS be a hockeystick.
Do you agree with this? If not, are you going to address it, audit it and have it corrected?”
Please note that Michael Jankowski continues the confusion introduced originally by Steve in mixing up PCs with RECONSTRUCTIONS by saying (posting 29):
“Does a “pronounced hockey stick-shaped PC1 over 99% of the time” in 10,000 repetitions satisfy your “always be a hockey stick” criteria?”
— no, of course it doesn’t — as I have argued a number of times, a PC is NOT the same as a reconstruction.
It would appear that Steve’s original spin on suggesting that PCs are equivalent to RECONSTRUCTIONS has now travelled far and wide. I’m sure that he is as pleased about this as he is reticent about answering my pertinent questions at the beginning of this posting.
As for Michael Ballantine’s rant (posting 30) about “But when you are given an open invitation to publish exactly what you claim was censored you ignore it and go back to whining about not getting information that is none of your business” — when I originally posted the comment that Steve subsequently censored, I was taking Steve at his word and assuming it WOULD NOT be censored — I therefore did not keep a copy, assuming it would appear on climateaudit. However, if Steve did keep a copy, then he is welcome to publish it. I can assure you that the only thing anyone might find offensive about it is the fact that some of you may not agree with it contents.
The weights in Mann’s 15th century reconstruction are such that about 90% of the variance is explained by the North American PC1 and the Gaspe series. So conclusions about the PC1 and Gaspe apply to the 15th century reconstruction. Mann’s reconstruction is not robust to the presence/absence of the PC4 in a centered calculation (and also to the presence/absence of the flawed bristlecone pine proxies, which are mined by the MBH98 procedure). I’ve not said that a PC is “equivalent” to a reconstruction. I think that our observations have been expressed quite clearly.
In the case of Gaspe, the form of cherrypicking is a little different than the PC1. Mann et al. changed their protocols to insert a hockey stick shaped series so that 15th century results would be lowered and then misrepresented the start date in the publication. In financial statements, if you treat one account differently, you would have to disclose it in a footnote and special treatment of one account would raise all kinds of red flags with auditors. This form of behavior is over and above the PC1 modeling. However, I’ve experimented with making selections from red noise (mentioned in connection with Jacoby) and I’m satisfied that you can produce hockey sticks by cherrypicking as well. I plan to write an article on this.
The regression module of MBH98 also accentuates hockey stick shaped series, especially in the 15th century. I have obviously made concerted efforts to obtain source code to reconcile these results. If an examination of Mann’s source code disproves any of my observations,I will amend any observations that require amending and send appropriate notice to pacificresearch.org.
I think that your priority should be to get Mann to archive his source code in a public location so that these matters can be determined once and for all.
The posting in question was Hunter continuing to seek personal information on John A. I had already expressed my position on this. I would classify the posting as tiresome rather than offensive, given that my position on this was asked and answered. Contrary to the implication above, I did not give Hunter an open invitation to continue this line of posting. I had provided notice that I was not going to continue to discuss John A. with Hunter endlessly and will not do so in the future and do not plan to argue endlessly about Hunter’s past post on this topic. I have not and will not censor posts discussing principal components.
I do not promise to attend to every comment on this list as it would be impossible for me to do so, as I’m trying to do other things as well. I do read every comment and try to respond to some comments. It’s a bit random what I respond to. I’ve already responded to a disproportionate number of Hunter’s comments.
Steve: It is ironic that the very first paragraph of your article at the top of this thread contains the word "stonewalling". I think that most people would regard your response (posting 45) to my simple questions in posting 26 (repeated in posting 44) as an excellent example of sonewalling. Let me try again.
The document “2005 Index of Leading Environmental Indicators” at http://www.pacificresearch.org, contains the clear and unambiguous statement: "Scientists have shown that the process used to generate Mann’s graph would generate the same result from any series of random numbers." I presume that this statement is an interpretation of your own findings, as you have described them to the public. I understand this statement to mean that, if I gave Mann a set of proxies constructed from random numbers, then the resulting RECONSTRUCTION would ALWAYS be a hockeystick. Now, I would like to know whether you agree with either of these two statements. This requites simply a "yes" or a "no" (or possibly a "I don’t know"). If it is either "no" or "I don’t know", then I would also like to know whether you intend to address this issue, audit the statement by the Pacific Research Institute, and have it corrected. Is this a lot to ask? Your credibility depends on your response.
Steve: I think that most people would regard my response #45 as answering your question. If anyone else besides you feels otherwise, I’ll look at this some more, but, as of now, I think that the question is asked and answered (see #45).
Steve: you are still, apparently, incapable of answering a simple "yes" or "no". However, I would interpret your response (posting 45), "however, I’ve experimented with making selections from red noise (mentioned in connection with Jacoby) and I’m satisfied that you can produce hockey sticks by cherrypicking as well", to indicate a very definite "NO" — i.e. if I gave Mann a set of proxies constructed from random numbers, then the resulting reconstruction would NOT always be a hockeystick". Of course, as is well known (and practised) in contrarian circles, you can produce almost any result you like by cherrypicking — but such "proofs" don’t prove anything.
To summarise, for those who are baffled by your continuing evasiveness: the statement by the Pacific Research Institute that "scientists have shown that the process used to generate Mann’s graph would generate the same result from any series of random numbers" is WRONG — and you apparently agree that it is WRONG. So, now Steve, are you going to do the honest thing and get the Pacific Research Institute to correct their statement, which is essentially just a "promotion" of their anti-warming case?
Steve: As you phrase it: "if I gave Mann a set of proxies constructed from random numbers, then the resulting reconstruction would NOT always be a hockeystick". I would be inclined to say that if you had given Mann a double-blind test with a set of red noise proxies that he would have produced a hockey stick using his methods, which include: the PC method, cherrypicking (e.g. Gaspe), deletion of series if that is "better for our purposes" and overweighting in his regression module.
You said: "you can produce almost any result you like by cherrypicking, but this proves nothing". My point exactly – Mann has proven nothing. Please convey that to your associates.
re #47
In addition to what Steve responded, you need to modify your ‘always produce a hockeystick’ to ‘almost always produce a hockeystick.’ Then he could answer yes. The method which tends to give hockeysticks is quite simple. Any largish grouping of randomly produced ‘proxies’ will have some with hockey-stickish sections at the sensitive point. This will automatically give them a high weighing when the calculations are done by the method of Mann98. But occasionally, especially if we’re not talking an unlimited number of ‘proxy’ series, there won’t be a hockeystick in the appropriate spot in any of them and thus the need for ‘almost’.
Steve: so, the truth of the Pacific Research Institute’s statement "scientists have shown that the process used to generate Mann’s graph would generate the same result from any series of random numbers" now relies entirely on an assumption of persistent cherrypicking by Mann et al. of ANY random input data that you give them.
Steve: I have edited this comment to delete sentences containing inappropriate terms like "dishonest" or "conspiracy", whether John Hunter is applying the term to someone else or to me and to delete a sentence containing more endless haranguing about John A. In the future, I will not take the time to edit out such terms from John Hunter’s posts, but will delete such posts.
As to the question, the answer is no.
Steve and Dave: Ah, now Dave has come up with a different story (posting 48), which presumably does not resort to Mann et al. having to cherrypick the data. Dave’s mechanism seems to depend on an adaptive weighting process inherent in the Mann et al. technique. Presumably either Steve or Dave have done a Monte Carlo experiment with the Mann et al. processing method (but WITHOUT any cherrypicking) and can statistically quantify what Dave means by "almost".
This of course begs two questions:
1. If the Mann et al. processing technique "almost always" satisfies the statement of the Pacific Research Institute, then why did Steve need to raise the issue of (malicious) cherrypicking at all?
2. If cherrypicking IS part of the story, then what on earth is the relevance of Dave’s "almost always" theory? If Mann et al. are going to cherrypick, then surely they would do a thorough job and ensure that the statement of the Pacific Research Institute is ALWAYS true.
Perhaps Steve and Dave should get together privately and produce a consistent and believable story.
John, re #49 and #50, first of all, the term "nearly always" is exactly what we said in our GRL article. We showed that the Mann method produced hockey stick shaped PC1s over 99% of the time from red noise. We did so by carrying out a Monte Carlo experiment. If you read our GRL paper, you can see this. You can also examine the source code for this which has been archived at the GRL FTP site.
The issue of cherrypicking was raised in our EE article in connection with the Gaspe tree ring series, which was prominently discussed in our EE article. In that article, we reported that Mann et al had carried out a unique extrapolation of the Gaspe series that had the effect of lowering 15th century results under their method. We also pointed out that they had misrepresented the start date of the Gaspe series in the original publication. As a result, no reviewer or reader without specifically crosschecking the WDCP data against the as-used data would not know of this unique cherrypicking extrapolation.
Steve
Re#44,
Sorry to confuse you. If PC1 comes out to be hockey shaped 99% of the time using Mann’s methodology and fed random noise, and PC1 is the “dominant pattern” according to MBH98 (PC1 was 38% of the North American variation in MBH98), then it seems very likely to me that the resulting reconstruction will be hockey shaped in nature.
Michael, actually Mann does not seem to be currently contesting this issue. At first, he denied that the PC method produced hockey sticks from red noise, but, after the point was acknowledged by von Storch, Zweiers,Hubert, he more or less conceded that this issue in comments to the Wall Street Journal, moving his defence to the claim that this erroneous method did not “matter”. He did not withdraw his previous denial at realclimate, however. Steve
Re. 51, 52 and 53: Steve and Michael seem to have come full circle and are again confusing PC1 with a reconstruction. I think realclimate accepts that the Mann et al. technique does generate a hockeystick-shaped PC1 from random data. But what they also say is that such patterns only explain a tiny fraction of the variance of the (random) input data — so OTHER PCs have to be included in the final reconstruction. (Michael’s posting 52 is of course rubbish: just because, with one particular data set, PC1 is the "dominant pattern", DOES NOT mean that PC1 is the dominant pattern with ANY (e.g. random) input data).
So Steve: I have been asking for this for a while. Please do a full Monte Carlo simulation using random input data and The Mann et al. procedures (WITHOUT any cherrypicking) and show us the statistics of the RECONSTRUCTIONS (i.e. how many look like hockeysticks?). Only once you have done this can we move on to the issue of cherrypicking.
Steve: Show me where realclimate has acknowledged that the Mann technique yields hockey stick shaped PC1s. Mann denied it in a response to a question from Steve Hemphill, so they need to come clean about it. Please read some of my postings on PC significance: if you put tech stocks in the data matrix, they are "significant" under a Presisendorfer test, that doesn’t make them temperature proxies. All of their arguments are just attempts to get bristlecone pines into the mix – where one series dominates the reconstruction.
We have shown that Mann’s reconstruction lacks statistical skill (R2 of 0.0); that he withheld the R2 statistic; that the reconstruction is not robust to the presence/absence of bristlecone pines; that MBH’s claims of robustness to all dendroclimatic indicators were a misrepresentation; that proxies essential to the hockey stick shape are flawed. Right now, Mann has not responded to any of these points, except with completely invalid arguments about Preisendorfer significance. I would like a little broader understanding of these points before introducing a new line of argument.
Right now my priority is to work on some of the other Hockey Team articles. I’m in the middle of several interesting projects.
If Mann makes his code public, I’ll probably change my schedules and do another round on MBH. I think that he’s eventually going to be forced to disclose his code, sooner rather than later, and I’d rather wait and do simulations using actual code. Steve
Steve: Another Hunter post contained the term “dishonest” and has been deleted.
Re #55. well, ‘McIntyre’ (being referred to by your surname, not nice is it?), we don’t know the context, who it was allegedly directed at, or the why of it. But, if the use of ‘dishonest’ is to be censored why did you allow John A (the ‘Mod’ here) to post this thread header – http://www.climateaudit.org/index.php?p=171 ? And why was he allowed to describe the sentence with the word ‘dishonesty’ in it as "… a zinger of a first sentence:"? Because of the letter ‘y’ or because of double standards?
Copy taken.
Steve: Peter, I was hoping not to have to try to develop and administer policies on flames. The standard that I used for Hunter was this: was this a form of expression that I would allow myself to use in making a posting about someone else? In this case, it wasn’t.
I can see some points of distinction between John A’s post and Hunter’s. First, the comment was from a public conference. Second, the quote was not personal. I can see arguments the other way as well.
I very much try to avoid making personal comments and to simply comment on the text. (You can perhaps find some inconsistencies, but it’s still something that I try to do and am pretty consistent about doing.) In some cases, text comments can be pretty strong. For example, I have stated quite clearly that Mann et al. have made various misrepresentations about their methods – that’s a comment on various texts and typically the record speaks for itself. No one has ever contradicted any such claim that I have made. I do not comment about whether those misrepresentations were intentional, accidental or malicious. Other people may choose to pursue such matters, but I do not have access to original drafts and documents and simply am not in a position to comment formally (although I have personal views).
For now, I have other things to do besides developing flame policies and will respond somewhat situationally. Now try posting something critical at realclimate and come back and tell me if you still have a complaint.
Steve
Re#54,
I gave you an example from Steve’s research. If you were familiar with M&M’s works, you would know that the NOAMER PC1 is the “most influential series on the hockey stick’s final shape” (you apparently read the take from the realclimate folks, so why not read M&M’s works to see the take from the other side?). If NOAMER PC1 is hockey-stick shaped 99% of the time when fed random noise, and if it is the most infludential series, how often do you think the reconstruction would also be hockey-stick shaped?
Michael (re. posting 57): Sorry, but we must be talking at cross-purposes. The experiment that I want Steve to perform is to feed the Mann et al. procedures with time series composed of random numbers. These will generate there own PCs, of which PC1 will have something like a “hockeystick” shape (I think we agree on this), although it will clearly only explain a very small part of the variance of the original random data. Now, “NOAMER” refers to the North American tree ring network, from which McIntyre and McKitrick (in their GRL paper) derived the autocorrelation function used to generate the random data in their Monte Carlo experiment. In this experiment, they found that the Mann et al. procedures mostly produced a hockeystick-shaped PC1.
I think you are confusing two meanings of “PC1”. In MBH 1999, the PC1 of the NOAMER network (called “ITRDB” by MBH) was used as a proxy record. In McIntyre and McKitrick’s GRL paper, they used the NOAMER proxy records (but NOT the NOAMER PC1) to derive autocorrelation functions from which to generate random data in their Monte Carlo experiment — in which they found that the Mann et al. procedures yielded a hockeystick-shaped PC1.
I therefore do not understand what you mean by “if NOAMER PC1 is hockey-stick shaped 99% of the time when fed random noise, and if it is the most influential series, how often do you think the reconstruction would also be hockey-stick shaped?”. What do you mean by “NOAMER PC1” (the proxy series used by MBH 1999, or the PC1 derived from the Mann et al. procedures using random data with the same autocorrelation function as the NOAMER proxies)? How do you “feed random noise” into the “NOAMER PC1”? Perhaps Steve can help here.
Perhaps you should note that the NOAMER proxies hardly have any bearing on the experiment I want Steve to perform — they could be used to provide the autocorrelation function from which the random input data is derived (as in McIntyre and McKitrick’s GRL paper), or, alternatively, they would not be used at all and the autocorrelation function would be derived from some other “reasonable” source.
As to your final question “how often do you think the reconstruction would also be hockey-stick shaped?”. My tentative answer would be “hardly ever”, but we won’t know for certain until Steve actually does the experiment.
***How do you “feed random noise” into the “NOAMER PC1″Ɀ***
You don’t – a typo/omission on my part. My statement, “…if NOAMER PC1 is hockey-stick shaped 99% of the time when fed random noise…” should have inserted the words “Mann’s algorithm” before the word “fed.”
***Perhaps you should note that the NOAMER proxies hardly have any bearing on the experiment I want Steve to perform ***
But they have very much to do with Mann’s reconstructions (as did the question I asked you – hard to tell if your response is related to Mann’s reconstructions or the experiment you want to see).
John Hunter. Will you please stop wasting everybody’s time asking for a Monte Carlo experiment of Mann’s procedure until such time as he makes his code public. In case you missed it, Steve has already indicated that he will probably do it when the code is available and will not waste time on the matter until then. Your question is therefore asked and answered. Many times.
From #54
“If Mann makes his code public, I’ll probably change my schedules and do another round on MBH. I think that he’s eventually going to be forced to disclose his code, sooner rather than later, and I’d rather wait and do simulations using actual code. Steve”
We may never know the answer to that question. To answer it correctly we would need to know the algorithm Mann applied. As we already know, the description of the algorithm used to process the North American tree network was woefully inadequate to allow a true replication, and seems that the same is true for the temperature reconstruction. As ever, with a complex piece of software, the only way to truly replicate the algorithm is to use the original source code. Unless Mann provides his source code, we will probably never know the answer to your question.
It seems you are happy to make demands on Steve to do the impossible (derive Mann’s full algorithm – undocumented quirks, coding errors and all – purely from his output data set and limited description) but unwilling to request Professor Mann carries out simple tasks (provide his source code, or a sufficiently detailed algorithm definition with comprehensive interim data sets to allow true replication).
My personal view of this question (that we may never find the answer to) is that I would expect the hockey stick shape to be unduly amplified in the final result.
To understand why, it is necessary to be aware of the reason and nature of the promotion of a hockey stick shape from the decentred EOF analysis. The promotion is caused because EOF analysis does not separate out the first and second moments of a data series. As such, the "variance explained" is only truly a variance (in the statistical meaning of the word) in the case of a centred EOF analysis, and should not be interpreted purely in this way for the decentred case. The offset introduced in the decentring process is integrated into the first principle component. The result from noise is essentially a step function, (step bandwidth constrained by the noise or underlying signal) with one average value outside the "training" region and another value within the "training" region. Circumstantial evidence, but this step function is imprinted on the temperature reconstruction.
But whatever the cause, the inability of EOF analysis to separate first and second moments should make it abundantly clear that the algorithm used by Mann not only "mines hockey sticks" but also amplifies the associated variance explained, due to the integration of the mean offset into it. (In truth, of course, it is the process of amplification of the variance explained that results in hockey stick being promoted to PC1). Note that the a
Spence, I apologize for the server cutting off the post. I’m trying to get it fixed, but have been frustrated so far. It’s ting me up as well. Steve
Michael, Michael and Spence_UK (postings 59, 60 and 61): this thread started with my posting 26, which noted the statement by the Pacific Research Institute that “scientists have shown that the process used to generate Mann’s graph would generate the same result from any series of random numbers”. This statement is presumably a misinterpretation of Steve’s work and I asked Steve to correct it if it was wrong. He has neither done so, nor justified the truth of the statement, sticking to the line that “I’d rather wait and do simulations using actual code”, which is similar to Spece_UK’s response, “we may never know the answer to that question ….. to answer it correctly we would need to know the algorithm Mann applied”.
So if you don’t know and if Steve himself doesn’t know, why do you guys allow the Pacific Research Institute to make this totally unsubstantiated allegation?
I think the word Steve would use is “promotion” …..
Steve, it was a bit of an essay – I really must stop doing that….
I kept a copy of the original. If it helps I could repost as two separate bits and we could delete the old one.
Re #46: At 4.55 pm blog time, I posted a comment on Pacific Research, explicitly stating that Ross had sent an email to Pacific Research pointing out that they had somewhat misinterpreted our findings. What possible objection can Hunter have to this? While it was very much my intention to leave MBH simulations alone for a while (I’m really trying to get some other work done) and it completely frustrates me that people like Hunter give Mann total licence to stonewall, I did the simulations in question and confirmed what had seemed obvious to me. I acknowledge that I can hardly object to checking what seems obvious – I do it all the time. In any event, the simulations confirmed that the PC results carried over to the reconstructions with some remarkable parallels. I’m not going to discuss this topic any more for a couple of weeks as I really have some other stuff that I’ve got to do, so please no more harangues on this for a while.
Steve (posting 65): you ask “What possible objection can Hunter have to this?”.
I don’t have any objection, except that I wonder why it has taken 8 days and many emails for you even start to address the issue I raised in posting 26. But thank you anyway.
Re#63,
What’s this “you guys” stuff? Some/many/most of the people who post here simply read, comment, and discuss topics every once in awhile and have no other relation to M&M. Being involved in a thread you started makes it our duty to interpret an “allegation” on another website the same as you do and then have them correct it as you see fit? FYI, you provided a link to Pacific Research but no direction to finding where that quote was located. Without specifics, it’s difficult to determine the full context of the statement and to determine if M&M are actually the ones being referred to as the scientists (if it’s not reference as such, then it’s an interesting conclusion you came to, since many Mann supporters seem to claim M&M aren’t scientists!). Maybe a more fine-tuned link next time (or at least chapter/page number?), or mention that M&M’s work was reference?
I don’t know if you interpretted their statement correctly, anyhow. I don’t think they’re even saying you’d “always” get a hockey-stick reconstruction if you fed random numbers into Mann’s methodology. It sounds more related to the statistical robustness.
Regardless, it looks like M&M should have made you happy in #65.
Re #63
Whilst I say we may never know the full, true and correct answer to the question, my post went on to explain exactly why I would expect the hockey stick shape to be incorrectly amplified in the temperature reconstruction (a result it would appear Steve has now confirmed). Unfortunately Steve would have to use what is as close as possible to Mann’s algorithm based on his inadequate description, but given how close Steve has got so far it is probably as good an answer as we can get.
As for the Pacific Research website, as far as I know Steve is not responsible for the content of that site, so I consider your “expectations” unreasonable. I’m sure Mann has not double checked every possible use of his hockey stick to ensure the t’s are correctly crossed and the i’s correctly dotted. The internet is a pretty big place and we have to accept that there are other things to do except check that nobody anywhere in the world is misquoting you at any point in time. Having raised the point, Steve has pointed out to them they need to slightly re-word their article. I don’t see what else you can expect him to do.
Comment #61 repost – explanation of why I would expect Mann’s algorithm to cause hockey stick temp reconstructions PART 1
####
We may never know the answer to that question. To answer it correctly we would need to know the algorithm Mann applied. As we already know, the description of the algorithm used to process the North American tree network was woefully inadequate to allow a true replication, and seems that the same is true for the temperature reconstruction. As ever, with a complex piece of software, the only way to truly replicate the algorithm is to use the original source code. Unless Mann provides his source code, we will probably never know the answer to your question.
It seems you are happy to make demands on Steve to do the impossible (derive Mann’s full algorithm – undocumented quirks, coding errors and all – purely from his output data set and limited description) but unwilling to request Professor Mann carries out simple tasks (provide his source code, or a sufficiently detailed algorithm definition with comprehensive interim data sets to allow true replication).
My personal view of this question (that we may never find the answer to) is that I would expect the hockey stick shape to be unduly amplified in the final result.
To understand why, it is necessary to be aware of the reason and nature of the promotion of a hockey stick shape from the decentred EOF analysis. The promotion is caused because EOF analysis does not separate out the first and second moments of a data series. As such, the “variance explained” is only truly a variance (in the statistical meaning of the word) in the case of a centred EOF analysis, and should not be interpreted purely in this way for the decentred case. The offset introduced in the decentring process is integrated into the first principle component. The result from noise is essentially a step function, (step bandwidth constrained by the noise or underlying signal) with one average value outside the “training” region and another value within the “training” region. Circumstantial evidence, but this step function is imprinted on the temperature reconstruction.
Comment #61 repost – explanation of why I would expect Mann’s algorithm to cause hockey stick temp reconstructions PART 2
####
But whatever the cause, the inability of EOF analysis to separate first and second moments should make it abundantly clear that the algorithm used by Mann not only “mines hockey sticks” but also amplifies the associated variance explained, due to the integration of the mean offset into it. (In truth, of course, it is the process of amplification of the variance explained that results in hockey stick being promoted to PC1). Note that the amplification does occur with pure noise, and the variance explained even with no signal is higher than one would expect for pure noise. The greater the underlying signal (of any type), the greater the potential amplification of the variance explained, for the reasons outlined above.
It is not unreasonable to assume that the amplification of variance explained in this case would have a material impact on the imprint of the “hockey stick” on the final temperature reconstruction. Could this have happened by accident? Could it be a form of subconcious data selection? Could it be a deliberate act? We’ll probably never know the answer. But the bottom line has got to be that the method applied has less validity, in this case, than a simple centred PCA. Health warnings almost always come with decentred PCA and this should be no surprise to someone with experience of them.
As an aside, it is important not to confuse M&M’s test (which is primarily of interpretation) for the test conducted by MBH (which is primarily of significance). The two are quite different, but relevant, experiments. The MBH test must be passed for a result to be significant; the M&M test must be passed for a result to be meaningful.
There are many other problems I personally have with the use of PCA for this task, such as the inversion and integration of inversely correlated temperature sets, which could also result in meaningless output. These are touched on in M&M but the full danger of this (IMHO) is not outlined. Combined with other criticisms (e.g., von Storch), I see no reason to have any faith in MBH98 or 99 as historical temperature reconstructions.
Michael Jankowski (posting 67): you ask “what’s this “you guys” stuff?”. Michael, it’s simply following the example demonstrated by Steve Mcintyre’s “people like you” (posting 12) and “people like Hunter” (posting 65). If you object to these kinds of generalisations, please direct your complaints to Steve.
Re #71, Incorrect Mr Hunter, it is simply your rude and obnoxious, petty, childish ways getting the better of you again. Please post something of substance on the data, or post your nitpicking on websites on all sides of the debate.
Re#71,
Posting 12’s “people like you” – I’m not sure exactly what Steve means by that, but I assume there are people like you?
Posting 65’s “people like Hunter” – Clearly refers to people who permit Mann to stonewall but expect others to do things on command. I can say that the latter certainly fits you. As to the former…care to comment? Steve seems to think it fits. Do you want Mann to release his source code? Does he receive the same “tedious prodding” that Steve got here?
I find both of those generalizations acceptable in the context of their posts. Regardless, they are completely irrelevant to why I raised the issue of the “you guys” reference you made.
What I didn’t find acceptable in #63 (that I referred to in #67) is that you specifically directed a post to myself and two other people saying that we should be correcting an error on the Pacific Research Institute’s website. I can’t speak for the other two people you were directing your post towards, but I have no control over the Pacific Research Institute or their website, no relation to M&M’s work, a different interpretation of the comment on the Pacific Research Institute’s website than you did, and had not been given any substantiation to show that the comment on the Pacific Research Institute’s website was in reference to M&M’s work in the first place as opposed to someone else’s!
Keeping all that in mind, can you explain why it is acceptable for you to question why someone like me hasn’t directed the Pacific Research Institute to change their website?
Re:#66
“I wonder why it has taken 8 days and many emails for you even start to address the issue I raised in posting 26.”
Really?! Eight days? I am shocked, totally shocked?
What is John Hunter’s opinion on Dr. Mann has responded or not responded to requests from his critics?
Re #72, err, who was it who referred to me as ‘Hearnden’ then Snack? (clue – http://www.climateaudit.org/index.php?p=176#comments ).
It seems you can sling it but when it’s returned you whine – for future reference how would you describe such behaviour? And, how would you react if I defended another poster here thus: ‘He does not have independent means and is worried about repercussions from people like you.’?
Steve and John, Knowing that you do not want to waste time coming up with a posting policy allow me to make one small suggestion that may clean things up a bit. It may be a little after the fact but how about starting a separate thread to move all comments that "discuss" or "refer to" the anonymous nature of any of the commenters on this site. Simply move them over to the other thread. Nothing censored out, just certain some comments moved to a more appropriate category reflecting their contents.
John replies: If anyone knows how to move comments from one thread to another in WordPress, please let me know.
Steve: I hope this can be figured out. It’s a great idea.
Mr Hearnden, what is your problem ? Where I am whining, I merely echo Mr Hunter’s little “bon mot” from earlier, and I note you didn’t leap up and object to that characterisation. Hypocrit.
I have given up challenging you to, just for once, actually discuss data instead of posting prissy little ad homs. Of late you have not upset my expectations.
Re#77 Ed, this place really has got poisonous. But, I’ll leave it at that rather than deliver the kind of reply your post deserves.
But who turned it poisonous Mr Hearden ? Your unending series of nitpicks and steadfast refusal to deal with the substance of the data being discussed hardly lends to amicable discussion. I am happy, at any time, to attempt to discuss the significance or otherwise of the various issues with the underlying data and methodology used in paleoclimate reconstructions, but you won’t engage. If you simply want to cheer on what the “experts” want you to think, why don’t you stay at Realclimate where such attitudes are welcome ? I treat these comments as an opportunity to consider and post (with deep understanding or not) on the underlying data. You don’t have to accept Steve’s (or any else’s) interpretations as long as you engage with the data.
But, Peter (to revert to your first name), you haven’t once in numerous posts ever seriously disputed the data ! You have cast doubts upon the people, their motivations, their expertise or competence, and their identities, and that has been your sole contribution. Do you wonder why we (or me at least) get frustrated when we see yet another non relevant Peter Hearnden post. Even now, you think my posting “poisonous”, did you make such a comment on Mr Hunter’s little effort earlier ? No, but why not ? And in response to the question in #75 re indentities, defend all you like, I have no objection, but defend his post, his data, his conclusions, not his personality.
Maybe you could create some sort of a spreadsheet/chart for data requests and status. That would show the amount of stonewalling…
One Trackback
[…] comments on my efforts with Jacoby not systematically, but in some anecdotes here , here , here and here . I will post up some details about Jacoby which are […]