As pointed out by an anonymouse troublemaker (ain't we all) Mann and Co have fired a late salvo at the end of the 2008 hurricane season in GRL
We obtain new insights into the reliability of long-term historical Atlantic tropical cyclone (‘TC’) counts through the use of a statistical model that relates variations in annual Atlantic TC counts to climate state variables. We find that the existence of a substantial undercount bias in late 19th through mid 20th century TC counts is inconsistent with the statistical relationship between TC counts and climate.Mann, Sabetelli and Neu find that the undercount is one (1.2 to be precise) by comparing the observed number of Atlantic tropical cyclones to that predicted using three proxys: sea surface temperature, El Nino and NAO indicies. The undercount bias is adjusted to give the best fit.
UPDATE: In the comments Henry Anonymouse asked for the full Fig. 1. (see below). By eyeball, it is pretty clear that the sea surface temperature is the driver, with El Nino and the NAO serving as modulating factors.
Maybe Pielke Jr. will run a rebuttal in Energy & Environment.
ReplyDeleteUPI already has a story on Mann's study.
Mus musculus anonymouse
Eli - can't get in to link for Mann & Co.
ReplyDeleteCould you show all four charts?
I'm assuming there is a chart for hurricane counts, SSt, El Nino, and NAO.
- Henry
Not sure that RPJr will run a rebuttal. Seeing as he's now using cyclones as an example of how people need to adapt to climate change, he obviously thinks that there's a connection.
ReplyDeleteThanks for the full charts.
ReplyDeleteAs you say, it does appear that SST is main driver, but that last hurricane bar looks low (since all three "factors" were high).
I've spent some time at weatherunderground.com, and have read several posts about the Nino/Nina effect on hurricanes. I'm assuming that the "nino3.4" chart includes the nina events, also.
One more thing about the chart: it appears that after 1930 or so, there were more "above average" years.
It would be interesting to see if the ACE follows the "number of hurricanes" graph.
I'm wondering if the arctic oscillation would look like, plotted for the same time frame. Apparently, the +/- arctic oscillation has an effect on trade winds, which could effect hurricane formation.
Another blog says the arctic oscillation is predicted to go into a large negative phase. Colder arctic air, colder weather for upper U.S.
- Henry
You see the ACE at wikipedia.
ReplyDeletehttp://en.wikipedia.org/wiki/Accumulated_Cyclone_Energy
Bocco
CA has an analysis of this paper. Apparently the M&S paper has hurricane activity as a leading indicator of SST, not the other way round.
ReplyDeletesee here http://www.climateaudit.org/?p=2463#comments
Does this matter? I'd like someone to comment
CA has an analysis of this paper. Apparently the M&S paper has hurricane activity as a leading indicator of SST, not the other way round."
ReplyDeleteActually, that's not quite what McIntyre claims.
You have to parse his words carefully:
"They regressed this year’s Atlantic storm count against the coming winter Nino and NAO."
The issue of SST is one that McIntyre does not address directly (curiously).
Sometimes it is not what one says but what one does not say that is most significant.
The CA crowd has it all wrong, TC count is not a leading indicator, it is a trailing indicator -- El Nino events have an influence on TC count (upwards) in the year following an El Nino event. McIntyre tortures an abiguous sentence in the paper to mean something it obviously does not mean. And of course the commenters at CA have clearly not read the paper.
ReplyDeleteRattus Norvegicus
ok thanks for correction on SST - my simplification, that's all.
ReplyDeleteRattus: yes we all thought that Nino was a leading indicator. However when McIntyre looked at the data, it became apparent that Mann processed it as a trailing indicator. McIntyre then looked for an explanation in the paper and found the ambiguous sentence.
Does it matter? Has McIntyre got it wrong?
Mr. Mcintyre needs to familiarize himself a bit with the science before spouting off nonsense. El Nino peaks in the Northern Hemisphere winter, but the event is already emerging (and having impacts) the preceeding spring. So, for example, the current La Nina is expected to peak within the next two months. This is the same La Nina that was believed to(and apparently did, slightly) lead to a relatively active season. This is due to a reduction in atmospheric wind shear, which is favorable for tropical cyclone formation. Note that the La Nina event, by convention, would be measured by the SOI or Nino indices averaged over the upcoming winter, but that same event indeed impacted this year's storm season (primarily the latter part of the season). There is a separate and essentially unrelated impact whereby El Ninos and La Ninas influence tropical Atlantic and Carribean SSTs in the following season (warmer the year after an El Nino event). Of course, as noted in the paper, that impact is subsumed by using those SSTs themselves as a predictor. It doesn't take a whole lot of research to learn all of this. McIntyre should do a bit of backround research and try to learn something first, before he makes a fool of himself.
ReplyDeleteActually, even Landsea agrees with Mann on this one:
ReplyDeleteExtremes in the ENSO cycle typically develop during summer, peak in late fall, and decay into the following spring. Therefore, for this analysis, we choose to define an ENSO year as the calendar year in which the El NiƱo develops.
http://www.aoml.noaa.gov/hrd/Landsea/elnino/index.html
McIntyre should do a bit of backround research and try to learn something first, before he makes a fool of himself.
ReplyDeleteWhy?
This will have no influence on his ability to try to impugn the people he doesn't like.
Best,
D
By the way, Roger Pielke Jr. gets a kind citation in today's story that ran in National Review Online.
ReplyDeleteMus musculus anonymouse
Somebody needs to contact Pielke Jr. and ask him what he thinks of this study.
ReplyDeletepielke@cires.colorado.edu
He's been ducking the discussion for too long.
I was merely trying to report the statistical model used in Mann and Sabatelli. It is my understanding that they regressed 1997 cyclone counts against 1997-1998 winter indices of NAO and Nino. I didn't comment on whether that was a good idea or a bad idea, merely that that was what they appeared to have done. I base this opinion not merely on the comment in the paper, which is not as clear as it might be, but on performing the regressions where I found that the claimed r2 statistics were obtained only by regressions against the following winter's indices.
ReplyDeleteOthers can discuss whether this is physically realistically or not.
However if this is actually their model, then any "prediction" of 2007 cyclone counts using their model requires knowledge of 2007-2008 winter NAO and Nino, which they do not discuss.
If their model uses prior year's NAO and Nino, then I have been unable to obtain the reported r2 using their data.
Steve McIntyre
Steve McIntyre:
ReplyDeleteI'm sure you have noticed (in fact, I know it), but you are getting a bit of "exposure" over Tamino's place.
Some guy posting as fred (might that be YOU?) is defending CA like there's no tomorrow, but is doing you no favors, since he is being called to the table on his bogus claims.
Just one example:
fred says: "you do not find accusations of fraud" [on CA]
That was quickly shown to be false with a link to a comment on CA that alleged fraud in the case of the hockey stick.
But then I don't need to tell you about that one because you removed the fraudulent charge immediately after it was exposed on Tamino's blog.
Nice try, but google cache still shows the claim of fraud that you allowed to stand for 3 weeks in the comments (comment #151) here
allegation of fraud made by Steven Mosher on Climate Audit
"I imagine the people who exposed the piltdown man were anti evolutionists?
"When you see the hockey stick, think Piltdown man. The fraud didn't make evolution
a false theory, but it did lead some down the wrong path for some time."
//end CA comment
Also, you state above with regard to the latest Mann paper that "I was merely trying to report the statistical model used in Mann and Sabatelli"
But your post on CA indicates a bit more:
The assignment of the 1997-1998 Nino to 1997 was not just for “simplicity”, but to improve the stats. It’s hard to imagine that they didn’t also do calculations using the preceding season indices, discarding these calculations when they didn’t work as well. This sort of data snooping needs to be reflected in confidence estimates, but isn’t done here."
Whom do you think you are trying to kid?
Apparently, you think everyone is as stupid as your cheer-leading squad at Climate Audit (you know, the ones who got you that Blog award)
More allegations of fraud in the comments at Climate Audit (and pretty close to it in the post)
ReplyDeleteThere were several to choose from in this particular thread, but here's my favorite , since it hits upon just about every major area of climate science:
/////////////
Beginning of quote
Comment 21
Lizi says:
June 27th, 2007 at 10:41 am
"After reading through the IPCC reviewer comments - and the STUNNING way the IPCC has just ignored every reviewer that doesn’t support their hoax, the only conclusion I can come to is that this is a cartel of very well organised fraudsters."
1) this current Briffa fraud
2) the previous Mann hockey stick fraud
3) the weather station UHI fraud
4) the “H2O postive feedback” fraud
5) add to the list…
This has gone beyond ‘questionable science’, beyond ‘incompetant science’, beyond ‘negligent science’. It`s time to start calling this what it is. Deliberate and pre-meditated fraud to defraud taxpayers of research funds and easy jobs."
"With Solar Cycle 24 now looking very mild, this will all be over within 2-3 years. Everyone knows global average temps have slightly cooled since 1998 and the way SC24 is looking, the cooling will probably continue for the next few years. Eventually, even the media propagandists will start to cover their tracks….and who are they going to blame - “The Team” et. al.
I wouldn’t be suprised if around 2010, we are discussing on this blog, the criminal trial of certain leaders of this fraud…."
Harsh words I know, and not the oppinion of CA webmasters. Just my thoughts."
END of quote
/////////////////
I especially love "fraud to defraud" -- as if there is some other kind?
of course, Steve McIntyre "discourages allegations of fraud", as we can clearly see with the strong admonition that he included after the above comment on his blog (not!).
He did have this to say further down the thread, however
(comment 33)
"Please find an appropriate thread for any comments that do not pertain to Briffa."
That was in response to this comment
"The IPCC will almost certainly blame any cooling on China’s industrialization and aerosols."
But hey, i can understand how he might not want to break up the train of the fraud comments with distractions about aerosols.
In preparation for Steve McIntyre's cleansing of the above page (a procedure that could become a full time job), here's the cached page.
That's funny anonymouse. I've read comments on RealClimate advocating criminal charges for merely holding an opposing opinion on GW. Is this your basic line too??
ReplyDelete". . . what if they were held civilly and criminally responsible for their positions on issues that affect the species well-being?"
"Corporations need to be brought under social control, not fiscal control.
"Charters need to be revoked."
"Corporations that do not display active responsibility to the future of the species need to be dismantled and their assets converted to systems that will demonstrate social responsibility."
No need for a cached page, it's still there. And I can find more like it if you like.
http://preview.tinyurl.com/2hwayj
Post #47
- Paul S
As usual, Paul, you are ignorant on this one.
ReplyDeleteIf you follow the conversation over at Tamino's, you will see that McIntyre has promised to clean out any allegations of fraud that are brought to his attention, which means that the reference to the regular page will then no longer show the original allegation of fraud (which makes it especially confusing to slow people like yourself).
But then again, I would not expect you to understand that based on your other comments here.
Cheers at Climate Audit!
By the way, Paul
ReplyDeleteHow are those surface station post card sales coming along?
I know you have spent a lot of time defending the project (here and elsewhere) and just wanted to see how things were going.
Find any new barbecues? Burn barrels? Portable heaters?
I bet you have quite a collection of post cards by now.
Maybe you can send them out for Christmas.
Double Cheers at Climate Audit!
PS Don't let John V get you down.
Climate Science Hoaxus Poaxus
ReplyDeleteSteve, two points. You write;
ReplyDeleteThe assignment of the 1997-1998 Nino to 1997 was not just for “simplicity”, but to improve the stats. It’s hard to imagine that they didn’t also do calculations using the preceding season indices, discarding these calculations when they didn’t work as well. This sort of data snooping needs to be reflected in confidence estimates, but isn’t done here.
a) "not just for simpliticty"
Mann and Sabbatelli write;
For simplicity, the ‘year’ was defined to apply to the preceding storm season for both indices (e.g., the 1997/1998 El Nino and winter 1997/1998 NAO value were assigned the year 1997).
... they probably mean using NH winter values for NAO alongside the El Nino values for that same year, since monthly NAO indices are available throughout the year and commonly oscillate between postive and negative phase 2-5 times per year.
There is no need to imply dishonesty.
"but to improve the stats"
anonymous 7:13 describes a physical basis for the Mann & Sabbatelli choice.
You are mind-reading.
Which brings me onto...
This sort of data snooping needs to be reflected in confidence estimates, but isn’t done here.
I am not well versed in statistics, so can you explain this? Since the lagged and ensuing TC counts are two different populations, described differently by the physical model, are both expected to correlate to a given Nino index, according to the physical model/hypothesis, and since this model also suggests the correlation with the lagged count should be greater (see anonymous again), and since this is exactly what Mann and Sabbatelli plus yourself find, how exactly is this meant to effect (implied decrease) confidence estimates? IOW it is a substanitally different situation than fishing for significance on the same data series from the same population by using multiple tests.
who made the following statement in a thread in which his commentators attack CA & Steve McIntyre for making scurrilous claims of fraud etc against Mann?
ReplyDeleteThe idea that global warming will be no more troublesome than catching a cold is criminally irresponsible.
http://tamino.wordpress.com/2007/11/26/i-gave-em-the-truth-they-thought-it-was-hell/#comment-9899
and who censored my comment with the same quote at same site?
pot meet kettle :)
Anon 12:45, you have gone off on a tangent, so allow me to pull you back:
ReplyDeleteThe issue being discussed on Open Mind is specifically "unsubstantiated allegations of fraud", which McIntyre himself has agreed are serious and has agreed to delete.
If you have a disagreement with this, perhaps you might tell us all why you believe it is OK to make unsubstantiated allegations of fraud.
Other than that, please do try to understand the issue before interjecting your two cents .
sure pal, and then you can tell us why it's ok to state in the same thread that sceptics of AGW alarmism are 'criminally irresponsible"
ReplyDeleteAnother quote by tamino:
ReplyDeleteThe suggestion that money we don’t spend to mitigate global warming will somehow go to help ameliorate human problems is the kind of detestable lie that serves only the greedy.
I think I get it now. It's perfectly acceptable for tamino himself to make unfounded allegations of being "criminally irresponsible" and spreading a "detestable lie".
Also, the odious use of the term "deniers" by AGW supporters with its obvious connection to Holocaust terminology is no accident. Jim Hansen has even managed to bring in the Holocaust analogy in reference to trains hauling coal!
In testimony before the Iowa Utilities Board, here's what Jim Hansen had to say:
"I am testifying here as a private citizen, a resident
of Kintnersville, Pennsylvania on behalf of the planet, of life on Earth, including all species."
How's that for modesty?
He goes on to say:
"If we cannot stop the
building of more coal-fired power plants, those coal trains will be death trains – no less
gruesome than if they were boxcars headed to crematoria, loaded with uncountable irreplaceable
species.”
http://www.columbia.edu/~jeh1/NMAletters_20071121.pdf
If CA is going to be held to account, I trust we can rely on objections to this outrageous abuse of language also.
-Paul S
anyway, back to the Hurricane count and Mann paper
ReplyDeletegraphing named storms, especially when the NHC seems happy to name the merest puff of wind a storm, is misleading
let's look at ACE indexes and hurricane days shall we?
http://www.coaps.fsu.edu/~maue/tropical/hdays.jpg
http://www.coaps.fsu.edu/~maue/tropical/e_a_1970.jpg
yep, looks like those SSTs are the driver of lowest ACE index since 1970 all right
Paul S:
ReplyDeleteHow are those surface station post card sales coming along?
Here's another tidbit of gratuitous slander by Jim Hansen. But he's a prominent climate scientist, so it doesn't count, right?
ReplyDelete"The real deal is this: the ‘royalty’ controlling the court, the ones with the power, . . . are the captains of industry, CEOs in fossil fuel companies such as EXXON/Mobil, automobile manufacturers, utilities, all of the leaders who have placed short-term profit above the fate of the planet and the well-being of our children."
http://www.columbia.edu/~jeh1/distro_realdeal.16aug20074.pdf
Hey Jim, don't slander just a few people in those industries, slander them ALL. Climate scientists, especially yourself, must be given a free pass when making these kinds of unsubstantiated statements.
- Paul S
Wow, Paul I can almost feel your hatred for Hansen coming through in your comment.
ReplyDeleteYou should really get some professional help to deal with that. It's not healthy.
We all know such hatred for Hansen is why he gets attacked on Climate Audit so frequently.
But as you correctly point out, he's a scientist -- and the jesters he refuses to joust with are not.
BTW, how are those post card sales, Paul?
Paul S.
ReplyDeleteMaybe you don't quite get it. There's nothing wrong with calling someone a fraud....when they are a fraud.
So when people call McIntyre and crowd a fraud, they aren't being mendacious, just accurate.
And slander is only slander when it's not true. Michael Jackson can't sue you for calling him a child molester.
Oh, and how do I order the post cards?
Mus musculus anonymouse
Anonymouse, your hypocrisy is crystal clear. Don't worry, we get it.
ReplyDelete- Paul S
Ammann and Wahl confirmed the hockey stick results of Mann (see below) and McIntyre is now trying desperately over at Open Mind to paint his own alternative picture of reality.
ReplyDeleteROBUSTNESS OF THE MANN, BRADLEY, HUGHES RECONSTRUCTION OF
NORTHERN HEMISPHERE SURFACE TEMPERATURES:
EXAMINATION OF CRITICISMS BASED ON THE NATURE AND PROCESSING OF
PROXY CLIMATE EVIDENCE, EUGENE R. WAHL and CASPAR M. AMMANN
"The results presented here show no evidence for removing the MBH Northern Hemisphere temperature reconstruction from the list of important climate reconstructions of the past six centuries, on the basis of alleged “flaws” in its use of proxy data or underlying methodology. Indeed, our analyses act as an overall indication of the robustness of the MBH reconstruction to a variety of issues raised concerning its methods of assimilating proxy data, and also to two significant simplifications of the MBH method that we have introduced. The shape
of a single-bladed “hockey stick”-like evolution of Northern Hemisphere temperature over the
last 600 years is strongly confirmed within the MBH reconstruction framework (general
algorithm and proxy data)."
Perhaps we are simply to accept on faith that it's all just one giant conspiracy?
The whole thing is too kooky for words.
At Open Mind, Steve McIntyre said this about 'robustness' of the MBH98 results to inclusion and exclusion of tree-ring proxies;
ReplyDeleteIn response to two points by Chris O’Neill, You say that “Mann was only making the robustness claim in the context of data after 1750″. That’s simply untrue. The statement in this article was, as I quoted:
MBH98 found through statistical proxy network sensitivity estimates that skillful NH reconstructions were possible without using any dendroclimatic data, with results that were quite similar to those shown by MBH98 based on the full multiproxy network (with dendroclimatic indicators) if no dendroclimatic indicators were used at all.
The claim contains no restriction to post-1750 data. If Mann was aware that the statement was not true for the AD1400 step (and there is evidence that this was the case from the CENSORED directory), then, in my opinion, he had an obligation to disclose both the “good” results and the “bad” results.
Now if one has a 'large' dataset but the result depends on a small subset of data then one may have overestimated, if one does not check for robustness.
But if the dataset is 'sparse' and the sensitivity being tested is to the inclusion/exclusion of the bulk of the data, then it is to be expected such a test would fail. Why do such a test in the first place? In the AD1400 step of MBH98, they had already adjusted confidence appropriate to the sparsity of data stretching back. Steve is claiming 'sensitivity' should a) be tested for and b) (maybe) that it would effect confidence.
Like with the 'confidence' claim above, I don't understand.
Eli... anyone... can you help me out?
(by the way, Steve's claim You say that “Mann was only making the robustness claim in the context of data after 1750″. That’s simply untrue." is itself untrue, as the context of Mann's claim is apparent in the graph which directly follows and in a statement further down the page... as Chris O'Neill pointed out on Open Mind).
"one may have overestimated"... confidence.
ReplyDeletewhy don't you be a big boy and go post your thoughts at CA?
ReplyDeleteof course you will have to leave your little nanny blog protection at Tamino
Anon3:25
ReplyDeleteI don't know to whom you are responding. But since your demeanour suggests you are a rather aggressive posturer you can stuff your suggestion.
it's not about being 'big'
but getting results which are useful.
I taught myself PCA just to find out the 'broken stick' claim was a load of baloney. Now Steve is making claims related to 'confidence' I do not understand. I don't know if he's right, or not, but I've seen too much on CA which I know to be false, or exaggerated, or noise and posturing, to waste my time asking there and dealing with folk -- like you.
Meanwhile the folks on Open Mind and Rabbett Run are genuinely helpful.
Actually, one does not have to know anything about PCA to judge whether the hockey stick is real.
ReplyDeleteAll one needs to do is look with one's own eyes.
As Eli put it, It writes itself
McIntyre masturbates endlessly with his stats, but it means nothing because the hockey stick shape does not depend on PCA.
And McIntyre certainly did himself no favors by commenting on Tamino's blog.
It's true what they say about letting someone talk long enough (or giving them enough rope..)
Actually, one does not have to know anything about PCA to judge whether the hockey stick is real.
ReplyDeleteIndeed, but it helps to understand why.
because the hockey stick shape does not depend on PCA.
In fact the point at where my confidence in CA broke was when Steve altered Mann's purpose of doing a reconstruction without PCA from, showing that the hockey stick was not an artifact of PCA, into, 'accepting the hockey stick is an artifact of PCA, and trying to salvage a hockey stick from the data by other means.'
... that is a simple error in logic.
so you learned up on PCAs did ya?
ReplyDeleteread this, all of it
http://www.uoguelph.ca/~rmckitri/research/StupakResponse.pdf
Anon 6:28 said "So you learned up on PCAs did ya?"
ReplyDeleteIt's not "PCAs"
It's "PCA".
At least learn the lingo if you don't want to look like a complete idiot.
read the Wegman answers, you are the one looking like an idiot
ReplyDeleteif you think Wegman got it wrong, explain why.
The issue is not whether Wegman got it wrong or right.
ReplyDeleteThe issue is whether the hockey stick shape depends on PCA.
And it does not.
The fact that you believe it does is a further indication that you have no clue what PCA is all about (which was made perfectly clear at the start from your reference to "PCAs", at any rate)
I suggest you do a little research on "PCA." It's pretty simple to understand, actually. Doesn't take an Einstein (or even Wegman), at any rate.
so you learned up on PCAs did ya?
ReplyDeleteread this, all of it
Erm, no, but if you wish to be helpful then extract the bits that you think are relevant that were not in Wegman's report, which I have read and was somewhat disappointed by, since one half was a reiteration of M&M's claims and saying 'I agree', the other half was innuendo... I mean... social network analysis, concluding that statisticians ought be given more employment.
Given the letters after his name, I had expected more and better.
As anon8:19 stated, it's about understanding PCA. Not copying formulas out of a textbook or following convention, but actually understanding what the process is doing and for what end it is intended for, and how those two mesh together, or not.
The issue is whether the hockey stick shape depends on PCA.
ReplyDeleteno dummy, read the cite
if you wish to be helpful then extract the bits that you think are relevant
sorry, no time to educate morons
read the whole cite i gave above and then tell us how you are smarter than Dr Wegman
"no time to educate morons"
ReplyDeleteYou mean on PCAs? [sic]
If it quacks like a duck...
If it quacks like a duck...
ReplyDeleteyes dummy, focus on the intentional sarcastic typo instead of telling the world how you fail to comprehend Wegman's criticism
for less challenged readers, here is the cite that doofus refuses to comment on, it's quite clear
http://www.uoguelph.ca/~rmckitri/research/StupakResponse
"intentional sarcastic typo"
ReplyDeleteFirst rule of holes: when you're in one, stop digging.
Quack, quack, quack..
Quaaaak, quack, quack..
By the way, my Dear Mr. Quack:
ReplyDeleteThere are these things called "html tags" that allow you to embed those "link" thingies directly in your blog comments.
You might have seen them: the ones that take you directly to the site when you click on them (instead of requiring your reader to copy and past the URL into the browser box, which is sooooo 90's)
This tells you how to do it.
Even though you'll still sound like a quack if you keep writing "PCAs", at least if you learn how to do the links, you might look like a 21st century quack.
Glad I could be of help.
Without attempting to describe the technical detail, the bottom line is that, in the MBH original, the hockey stick emerged in PC1 from the bristlecone/foxtail pines. If one centers the data properly the hockey
ReplyDeletestick does not emerge until PC4. Thus, a substantial change in strategy is required in the MBH reconstruction in order to achieve the hockey stick, a
strategy which was specifically eschewed in MBH. In Wahl and Ammann's own words, the centering does significantly affect the results.
[snip]
Wahl and Ammann reject this criticism of MM based on the fact that if one adds enough principal components back into the proxy, one obtains the hockey stick shape again. This is precisely the point of contention. It is a point we made in our testimony and that Wahl and Ammann make as well. A cardinal rule of statistical inference is that the method of analysis must be decided before looking at the data. The rules and strategy of analysis cannot be changed in order to obtain the desired result.
Is he serious? In future, when a research project spends a year gathering data and moves onto analysis, and decide to summarize / represent / compress data series by PCA, and they estimate that say the three first PCs might suffice, and test the results against the full dataset, and find they left out far too much data. Well then they pack up tools and go home.
Or anyone wishing to use PCA decides a-priori to use the first 20 PCs to be on the safe side of Dr. Wegman's rulez.
Or they pack up and go home.
Suppost there is a digital image I wish to do some stats analysis on but the size makes it computationally expensive. I compress the image by PCA, thinking I might need to retain the first PC. The compressed image however after retaining the first PC looks like the
href="http://www.ia.pw.edu.pl/~wkasprza/researchPCA.html">butterfly butterfly in the top-left panel. Well, I can't change my mind, because the rules and strategy of analysis cannot be changed. So I proceed with the flawed analysis or pack up tools and go home.
PCA is a tool... it is used for a variety of reasons, ends, so the way in which it is used also varies.
PCs are a rotation of the data, that is all. The axes change, not the data. No 'artificial' 'generation' of 'artifacts' involved.
What was the purpose behind MBH98 use of PCA?
How does this differ from the butterfly example?
What are artifacts in the series of butterfly images, and how were they
created?
Dr Wegman's approach mirrors the methods of the parrot-education system.
That is to create a series of 'rules' to get the 'right' answer, that works in the classroom... one gets a tick or an 'A' by following the rules correctly, or copying them out of a textbook, regardless of how well the rules match reality, or even how useful they are in achieving ends. The strange rules here are that one may not estimate whether one has included sufficient PCs by comparing with the full data series, but one may do so by various forms of guesstimates before hand, in other words one may only estimate with one's hand tied behind one's back -- the correct decision process is the worse one.
It doesn't work in the real world.
It doesn't even make sense.
Of course the "rules and strategy of analysis" changed when M&M changed the centering method. For who-knows-what reason that doesn't 'count'.
His logic is shot to pieces.
So probably is yours.
That is possibly a result of spending too much time on 'rules', and theoretical work, than practical effort in aid of real goals.
More from Dr Wegman commenting on Wahl and Amman (2006).
ReplyDelete[WA}: "Second a related area of scrutiny of the MBH reconstruction technique arises from an atmosphere-ocean general circulation model (AOGCM) study, which
also examines the potential loss of amplitude [in the MWP] in the MBH method (and other proxy/instrumental reconstructions that calibrate by using least squares projections of the proxy vectors onto a single- or multi-dimensional surface determined by either the instrumental data or its [their] eigenvectors." Wahl and Ammann (2006, p.4 in the 24 February 2006 draft)
[Wegman] Again the MBH reconstructions do not correlate well with the model-based methods. Wahl and Amman (2006) offer the following explanation.
[WA] "However, a number of issues specific to the modeling
situation could arise in this context, including: how realistically the AOGCM is able to reproduce the real world patterns of variability and how they respond to various forcings; the magnitude of forcings and the sensitivity of the model that determine the magnitude of temperature fluctuations; and the extent to which the model was sampled with the same richness of information that is contained in the proxy records (not only temperature records, but series that correlate well with the primary patterns of variability including, for example, precipitation in particular seasons." Wahl and Ammann, (2006, p.5 in the 24 February 2006 draft)
The next sentence in the same paragraph reads;
In addition, the MBH method itself was inaccurately implemented by von Storch et al., leading to erroneously high amplitude losses (Wahl et al., accepted).
I guess he just forgot to quote this part of WA's "explanation".
Errors in logic.
ReplyDeleteIf the variance is artificially increased by decentering
Any choice of point from which variance is measured can be described as arbitrary or 'artificial'.
The MBH98 methodology puts undue emphasis on those proxies that do exhibit the hockey-stick shape
And what exactly is 'undue' about it in relation to the physical world?
Indeed, it is not clear that the hockey-stick shape is even a temperature signal because all the confounding variables have not been removed.
What is the purpose in doing a CFR?
Why does reconstruction skill drop when the 'hockey stick' shape is removed?
[Bart Stupak] Did you analyze this work by Wahl and Ammann prior to sending your final report to the Committee on Energy and Commerce? If so, why does
your report not alert the reader that these researchers had conducted a reanalysis of the MBH98 that corrected the only statistical methodology error discussed in the "Finding" section of your report and that these researchers found that recentering the data did not significantly affect the results reported in the MBH98 paper?
{Wegman] The Wahl and Ammann paper came to our attention relatively late in our deliberations, but was considered by us. Some immediate thoughts we had on Wahl and Ammann was that Dr. Mann lists himself as a Ph.D. coadvisor to Dr. Ammann on his resume. As I testified in the second hearing, the work of Dr. Ammann can hardly be thought to be an unbiased independent report.
Innuendo / slur / mind-reading / guilt presumed by association.
PS
ReplyDeleteIn other words, the hockey stick shape must be in the data to start with or the CFR methodology would not pick it up.
But Steve McIntyre says it's an artifact... or is that just sloppy language?
As I indicated above, the hockey stick shape does not depend on PCA.
ReplyDeleteThe PCA brouhaha is just a red herring.
M&M's criticisms were more than adequately addressed here
"We further show that the entire issue raised by MM regarding the centering convention used in PCA is spurious by demonstrating that similar results are produced whether or not proxy networks are represented using PCA at all".