Friday
Apr272012
by 
Bishop Hill
Bishop Hill Hockey Stick Illusion denial
Apr 27, 2012 Climate: HSI I came across this review of Michael Mann's book in Times Higher Education. The author, Jon Turney, is a green science writer, so you know exactly what to expect, but the thought struck me that it is completely amazing that Mike Hulme's is still the only review of Mann's book to even mention the Hockey Stick Illusion.
Are they all in denial or something?
Reader Comments (68)
And the aha moment, probably from St Andrews following the debate.
Correction (and addition):
none of these related reconstructions can establish that the modern warming anyhow may be seen as unprecedented, that it is indeed warmer than 1000 years ago, or that warming now is faster than anytime before. None!
And none of such plotted reconstructions centered within their error bars in light grey (if any) correctly assess and calculate confidence intervals for the presented estimated metric (global- or hemisphere averaged temperature), in the sense that the true metric lies within the presented bounds. Those error bars show something very different.
Go, Jonas!
I keep telling people this and they come back with the BBC bullshit
In the name of nailing this citation metrics drivel once and for all, Andrew Wakefield's seminal (sic) paper on Ileal-lymphoid-nodular hyperplasia, non-specific colitis, and pervasive developmental disorder in children has an astonishing 1449 citations, which comfortably thrashes Mann's 1998 and 1999 acts of deceit.
In addition to this astonishing list of citations, it also has the word 'RETRACTED' alongside it.
Mann's work has been repeatedly shown to be at least as flawed as Wakefield's. I am confident that in due course, his drivellings will have precisely the same endorsement and his career will be similarly terminated.
Footnote for 'chris' - papers are also cited by those who are lambasting the contents of execrable publications- metrics don't count 'for' or 'against'.
They also don't count how many times an author cites his own work. Mann et al consist of a group of around 40, though more reasonably about a dozen, authors that repeatedly cite their own work in every publication they submit. Not a surprise given the "climate community," particularly that relating to dendrochronology, is rather small. Too bad none of them actually bothered to take any real signal analysis coursework.
Mark
Sometimes I find that one can learn far more from what a writer does not say than from what s/he does.
Mann's latest opus - based on my own reading of the freebie sample and the reviews of those I trust - would have been more appropriately titled Portrait of the Artist as an Aggrieved Mann: A Novel.
The conspicuous absence in Mann's exercise in creative writing of any mention of The Hockey Stick Illusion is far more telling than anything Mann might have written about it.
Clearly Mann recognizes that he is skating on very thin ice - and would very much prefer not to provide his readers with the opportunity to check the "arguments" he advances.
chris:
"I'm also curious about the dissing of Dr. Mann's work, when there isn't anything in the scientific literature that indicates there's significant problems with it."
"Wegman's paper in which he attempted to trash Dr. Mann's research practices was retracted by the publisher of Computational Statistics & Data Analysis for fraud (Wegman plagiarised others work)."
chris,
Do you find that your arguments work with some people? I admire your craftsmanship, but I would question how sustainable it is.
BitBucket ...
I've read that DeepClimate (DC) post several times, and it is as you say a criticism of the Wegman report, particularly that it used the same renderings as displayed in M&M, and that these weren't average samples. That's fair to point out. Unfortunately that anonymous blogger DC then goes on to bring in Wahl and Ammann to somehow rescue Mann's work, which actually was its sole purpose. Thus he tries the same talking point as so many others, viz 'confirmed independently' and totally misses the point.
The term 'fraudulent' however is completely inappropriate wrt to M&M. I'd be careful to label Mann's many attempts as that, but cannot exclude that he knowingly fudges things in order to distort and mislead. And speaking of shrill language: DC:s final paragraph quite clearly shows what that he wishes for:
And it is still not about M&M or their critique .. not even the general contents of the Wegman report.
And if you compare the actual substance of the critique with the common practices within 'the Team' publications and their reviews, or just how uncritically things from that side are echoed 'with abysmal substance and scholarship, without due diligence or competence etc' ..
.. one would get quite a few very good laughs, but also gain insight into how weak their cases actually are, if you go to the bottom of it. One even might subscribe to DCs final (only minimally altered) contention:
Ah, the citations game. Seemingly small things can be very revealing of a person's character. One of the Climategate1 emails that sticks in my mind is the one in which a certain academic openly stated that he would inflate the number of papers authored by Phil Jones (to support, if I recall correctly, the latter's membership of some group or other) by including a handful authored by another P Jones. Some might discern a pattern of splicing disparate datasets, and a somewhat cavalier attitude to reality. The fact that this man was not only willing to lie on an application form, but was so open about his intention to do so, is also rather telling.
DaveS
" if I recall correctly"
Give us a proper citation ;)
I was interested to see that the comments under this review are almost all negative, does anyone know if there is any factor that would cause bias toward the sceptic side? There are a few brief posts by true believers but they seem to be uninformed and empty of content. I don't think that the reviewer can still be unaware of The Hockey Stick Illusion, if he reads the comments that is, as it is mentioned frequently in the thread.
Hi Jonas N, thanks for reading the DC article.
I'm confused that you say that it is not about M&M. The code snippets given are from the M&M scripts, reused by Wegman et al (ftp://ftp.agu.org/apend/gl/2004GL021750/2004GL021750-script.final.txt).
I'm not an R coder, but I think I can follow where it says "#SAVE A SELECTION OF HOCKEY STICK SERIES" at around line 512 in the M&M script, that it really is sorting and selecting the top 1% of 'PC1s', those most resembling a hockey stick. This is described in the DC page - do you not share that understanding of the code?
I do share that understanding .. Are you seriously suggesting that this somehow makes Mann a better scientist or Mann's work better science!? How?
Hi Jonas, I wasn't really interested in Mann per-se, I was trying to understand what has been said before: that when fed red noise "Mann's method" produces hockey stick curves.
The DC article seems to say that the M&M scripts do indeed produce hockey sticks from red noise! When the generated curves and sorted and the top 1% are selected, they have a striking resemblance to hockey stick curves. Is this what is generally understood to be happening in the script? And do the Mann scripts perform these same steps (sorting and selecting)?
That should have read "When the generated curves ARE sorted and the top 1% are selected..."
Yes, that's correct: When fed with red noise, Mann's algorithm produces hockeysticks almost inevitably. Some of the worse examples where shown in a figure by M&M (and by Wegman in fig 4.4)
If you fed random no-trend red noise 10.000 times into Mann's algorithm, essentially every one (99%) of them returned hockeysticks (defined as at least 1-sigma difference between the entire series, and the 'calibration period'). Half of them pointing up, the other one down. An unbiased algorithm should result such hockeysticks only those times when the random data indeed happened to be a hockeystick (pointing up or down), in ~15% (pointing up <8% of the time) and with a 1.5 sigma hockey-stick only 0.1% of the times.
The Mann-algorithm-rendered hockeysticks had indexes between 1 (sigma) up to about ~2 (sigma) 99% of the time. With 73% above 1.5 sigma, with 21%even above 1.75 sigma. And some 0.2% above 2.0 sigma! When fed with red noise!
Average hockeystick index was about ~1.6 sigma, and the 'criticism' you refer to was that those examples shown had ~1.9 sigma, and thus were not showing the average of the errors introduced, but rather some of the even (but only slightly) worse results.
You must have misunderstood this, because Mann did not generate and select among 10.000 different proxy series. Rather he selected proxies he liked (with already existing hockeystick shapes), and fed them into his own-devised algorithm that would vastly magnify and blow up the slightest tendency of a hockeystick-like wiggle towards the end, to get his (in)famous graph. In the way, and with methods that M&M after many years of trying to access the code, data, and methodology finally managed to show.
In other words: The selection of proxies, their quality, the introduced bias etc, are not part of the M&M (or Wegman) critique. Only the algorithm to thereafter inflate the 'hockeystickiness' of the chosen proxies ...
Manns choice of proxies is an altogether different story. Not less interesting, but just not what M&M focused on.
I hope this made it a little clearer, BitBucket ...
Jonas, thanks for your detailed explanation. I must say that the only thing that is really clear to me now is how little I understand about statistics and the handling of complex data sets :-(
I have to admit that I naively thought (up until a few weeks ago) that to produce the necessary graphs from the proxy data, one plotted the data on a pice of paper (or Excel etc) and drew a line through it. How wrong I was! I'm going to have to learn something more about data handling and stats - I may be gone for some long time...
Thanks for your patience (also to TheBigYinJames and Martin A, if they are around).
Steve McIntyre, fighting sickness and exhaustion, is back with a powerful broadside at Mann's book. Many comments of exceptional sympathy and support.
http://climateaudit.org/2012/04/23/checking-in/