The Ghost of Present ClimateBall ™

Good news: Stephen G. McIntyre [0], whom I prefer to call “the Auditor” for obvious reasons, has started to “promote” (in a technical sense [1]) the concept of ClimateBall ™. Better yet, he distorts it by labeling an indefinite group of commenters with it; compare and contrast with what he did so many times to those gyrating in the extended circles of the Kyoto Flames [2]. Yet again, the Auditor celebrates the spirit of ClimateBall ™.

Here’s one instance of distortion where AT (hereafter Anders) is implicated:

From time to time, Anders of the ATTP blog has attempted to understand the dispute, but uncritically accepts ClimateBaller doctrine, as for example, his following comment at Brandon’s:

People, however, clearly interpret the results of MM05 as implying that random red noise typically produces hockey sticks, rather than random red noise sometimes (probably quite rarely) produces hockey sticks.

This is completely untrue. MM05 did not imply that Mannian PCs “typically” produced hockeysticks: it stated it.

http://climateaudit.org/2014/09/27/what-nick-stokes-wouldnt-show-you

While his usage clearly excludes himself from being a ClimateBaller, “paying due diligence” [3] to this excerpt suffices to show that the Auditor channels the ghost of present ClimateBall ™.

***

Let’s commit some facts.

0th, the Auditor uses “from time to time” without substantiating this marker.

1st, the Auditor fumbles his link under “ATTP blog.”

2nd, the Auditor abbreviates the name [4] of a blog without mentioning it first.

3rd, the Auditor judges Anders’ attempt to understand as being “uncritical”.

4th, the Auditor calls “ClimateBaller doctrine” an unidentified set of beliefs.

5th, the Auditor quotes, as an example, a claim related to the Climate Wars [5].

6th, ClimateBall ™ [6] is independent from any claim in these wars.

7th, this Auditor’s usage of “doctrine” [7] may look like a slur.

8th, the Auditor quotes Anders without citation to the correct editorial [8] at Brandon’s.

9th, Brandon uses that quote in at least two other editorials [9, 10].

10th, Tom Curtis has covered that episode on his blog [11].

11th, the Auditor calls a comment one sentence from Anders’ comment.

12th, the Auditor quoted the last sentence of a long paragraph.

13th, that paragraph reads:

As far as your criticism of MBH98 is concerned, I don’t dispute the issues. It may also be true the using the MBH98 data to produce the red noise is largely irrelevant. It does, however, seem odd – as a physicist – to see people claim to produce independent random red noise, but to do so using the data they’re trying to compare to. Maybe that illustrates my ignorance with respect to what actually happens here, but it still seems a little odd. What seems indisputable, though, is that the 10 hockey sticks presented in MM05 (one of the papers, you probably know which one) were not selected randomly from their sample of 10000. They were chosen to be most hockey-stick like. People, however, clearly interpret the results of MM05 as implying that random red noise typically produces hockey sticks, rather than random red noise sometimes (probably quite rarely) produces hockey sticks.

14th, this paragraph is a part of a long comment, with many more paragraphs.

15th, this comment is followed by other comments by Anders on that thread.

16th, the exchange between Brandon and Anders rejoices the ghost of ClimateBall ™ past.

17th, the sentence we emphasized contradicts the Auditor’s accusation of lack of criticality.

18th, Anders’ comment ends up with a plea that is not in the spirit of ClimateBall ™.

19th, the Auditor’s proposition “this is completely untrue” starts with the pronoun “this”

20th, this “this” could refer to the results of MM05b [12] or how people interpret them.

21st, the Auditor claims that “this” is untrue because “MM05” stated it.

22th, Brandon ridiculed [13] Anders for conflating MM05a [14] and MM05b; cf. above.

23th, there is one “typically” in MM05b, and it is unrelated to the purported statement.

24th, MM05b does not state that “Mannian PCs “typically” produced hockeysticks.”

25th, even if we could reconcile the Auditor’s misspecification with a paraphrase like “that’s what we said,” to say something does not always mean that we imply it [15].

26th, to be relevant in rebuttal, “that’s what we said” implies what was said in MM05b.

27th, the claim that Anders’ claim is “completely untrue” is untrue.

28th, the Auditor’s neglects that Anders’ claim was about how people interpret this result.

29th, the Auditor’s untrue claim is irrelevant to what claimed Anders.

In that episode alone, the Auditor misrepresents both Anders’s, who tries to stay above ClimateBall ™, and ClimateBall ™ itself. This should be enough to show that the Auditor is one of the fiercest ClimateBall ™ players around. Well played!

***

Let’s end this note by correcting a belief proffered by one of my favorite ClimateBall ™ players, Brandon, whom I nicknamed Chewbacca [16] for less obvious reasons, unnecessary to recall at the moment:

It’s a term stemming from “Climateball,” a word I believe was invented by the user willard. […] While I believe the term was originally conceived as a way of referring to people on the skeptical side of the debate, it fits people like Nick Stokes and Michael Mann quite well.

http://climateaudit.org/2014/09/27/what-nick-stokes-wouldnt-show-you/#comment-732945

As this very “user,” I will attest not having “originally conceived” ClimateBall ™ “as a way of referring to people on the skeptical side of the debate.” I use contrarian [17] for that. The word ClimateBall ™ refers to the moves played in hurly burlies like the very one, by whatever side there could be, as long as the spirit of present ClimateBall ™ is channeled, just like the Auditor does in our episode.

References

[0]: http://past.is/fP3EW (Business Week)
[1]: http://climateaudit.org/2005/07/09/conflict-of-interest-1/#comment-34420
[2]: http://neverendingaudit.tumblr.com/tagged/thekyotoflames
[3]: http://climateaudit.org/2005/02/14/some-thoughts-on-disclosure-and-due-diligence-in-climate-science/
[4]: http://climateaudit.org/2008/09/22/say-my-name/
[5]: http://en.wikipedia.org/wiki/The_Hockey_Stick_and_the_Climate_Wars
[6]: http://neverendingaudit.tumblr.com/tagged/climateball
[7]: http://en.wikipedia.org/wiki/Doctrine
[8]: http://hiizuru.wordpress.com/2013/12/12/basic-truths/#comment-15
[9]: http://hiizuru.wordpress.com/2014/02/08/my-mistake-corrected/
[10]: http://hiizuru.wordpress.com/2014/08/06/basic-truths-part-two/
[11]: http://bybrisbanewaters.blogspot.com/2014/08/thoughts-on-mcintyre-and-mckitrick-05.html
[12]: http://climateaudit.files.wordpress.com/2009/12/mcintyre-grl-2005.pdf
[13]: http://hiizuru.wordpress.com/2013/12/12/basic-truths
[14]: http://multi-science.metapress.com/content/r27321306377t46n/?genre=article&id=doi%3a10.1260%2f095830503322793632
[15]: http://en.wikipedia.org/wiki/Irony
[16]: http://neverendingaudit.tumblr.com/tagged/chewbacca
[17]: http://neverendingaudit.tumblr.com/post/53053169520

Advertisements

About Willard

neverendingaudit.tumblr.com
This entry was posted in ClimateBall, Steven McIntyre and tagged , , . Bookmark the permalink.

182 Responses to The Ghost of Present ClimateBall ™

  1. Steven Mosher says:

    [Mod: This comment has been removed by the moderator]

  2. Joshua says:

    Brandon Shollenberger
    Posted Sep 28, 2014 at 3:46 PM | Permalink | Reply

    […]

    While I believe the term was originally conceived as a way of referring to people on the skeptical side of the debate, it fits people like Nick Stokes and Michael Mann quite well.

    I love Brandon.

    Ya’ just gotta love Brandon.

  3. Added the citation to Brandon’s comment for you to cherish it in its proper context, Joshua:

    http://climateaudit.org/2014/09/27/what-nick-stokes-wouldnt-show-you/#comment-732945

  4. Joshua says:

    Also quite beautiful:

    “Anders of the ATTP blog has attempted to understand …, as for example, his following comment at Brandon’s…”

    “People, however, clearly interpret the results of MM05 as implying that random red noise typically produces hockey sticks, rather than random red noise sometimes (probably quite rarely) produces hockey sticks.”

    “This is completely untrue.” MM05 did not imply that Mannian PCs “typically” produced hockeysticks: it stated it.

    Follow the bouncing ball, and then answer the question – what is “this?”

    You’d think for all the accolades he receives, Steve would have better than Triple-A skills.

  5. Michael 2 says:

    I am reminded of a small book, “Games People Play”:

    “The second half of the book catalogues a series of ‘mind games’ in which people interact through a patterned and predictable series of ‘transactions’ which are superficially plausible (that is, they may appear normal to bystanders or even to the people involved), but which actually conceal motivations, include private significance to the parties involved, and lead to a well-defined predictable outcome, usually counterproductive.”

    http://en.wikipedia.org/wiki/Games_People_Play_%28book%29

    All games have rules and structure. The covert game is deciding who establishes the rules — there’s a subtle but powerful shift in social power to the person that succeeds in defining the rules.

    I read a science fiction book by Barry Longyear many years ago, “The Tomorrow Testament” if I remember right. In it a human girl is taken by the aliens and raised in their culture to try to create a type of “bridge” between human and this alien culture. It includes some psychology and philosophy that I found interesting and which probably had roots in the Vietnam war (it’s anti-war and explores why war exists and why it is so difficult to extricate oneself therefrom).

    Anyway, the girl is in a classroom and the task is to create a game. Each child must obey all previously defined rules and create a new rule. Needless to say it becomes increasingly difficult. This human girl was last in line. She declared, “I win!”

    The rest of the class was astonished and angry, saying, “How can you say that?” and she said, “I created a rule that says I win.” The others said, “Any of us could have done the same thing!” to which she replied, “Yes, but I did it FIRST”.

    Anyway, she identifies the “binding rule” that has brought humans and aliens to war; each has a colony on a certain planet, and those colonists have invoked their sponsors into a war. The girl, now a woman ambassador, advises both sides to abrogate the treaty (Americans were treaty-bound to help the French in Vietnam). So, they do, and the war ends everywhere except on that one planet, which by no longer having the backing of their respective empires, must make peace or go extinct. So they make peace.

    And so it is that a generation later my niece spent some days in Hanoi and Ha Long Bay, one of the most beautiful places on Earth.

  6. Tom Curtis says:

    Willard, greatly enjoyed the post.

    As a slight update on my blog account to which you link, while playing around with the M&M05 cherry picked data, I tried some variant Hockey Stick Indexes. I did so because I noticed that a high HSI was not well correlated with having a positive slope in the 1902-1980 period. From memory (all I have to go one thanks to the 2nd Law of thermodynamics winning an argument with my computer), around one sixth to one third of the cherry picked high HSI pseudo reconstructions had a negative slope over that interval. I became determined to get a better index when I realized that any straight line with positive slope, in the absence of noise, got a high HSI based on the M&M05 index.

    I tried several indices, including correlation over the 1902-1980 period, ratio of standard deviations over the 1902-1980 period relative to the period to 1902, and closeness comparison of the angle of the trends to 1902 and from 1902. (I played around with the pivot year to find if that made much difference.) On average, the cherry picked highy M&M05 HSI pseudo-reconstructions performed poorly on all of these measures. Sufficiently so that the MBH98 and 99 reconstructions were picked out as performing better by a statistically significant amount. The key reason was that they all showed a declining trend till about 1890, whereas the cherry picked M&M05 examples almost universally jumped in mean value around 1800 or earlier.

    The upshot is that the results of M&M05 as regards to the ease with which short centered algorithms reproduce hockey sticks is not reproducible using other, more intuitive HSIs, and that are only obtained in the first place due to the extremely non-discriminating nature of the HSI used in the paper.

  7. In the hope of repeating past ClimateballTM games, there are some things in the quote that Willard uses that are wrong, or that I would have said differently had I known then what I know now. That, however, doesn’t change the point that Willard is illustrating. The bit that Steve quotes was referring to how people interpret the paper, not what the paper says itself.

  8. John Mashey says:

    Well, I generally too involved in other pursuits to play ClimateBall, but perhaps there was confsuion because:
    a) MM05 wasn’t necessarily the easiest to read for a general audience

    b) But the Jan 27, 2005 Backgrounder was quite clear:
    “We showed that the PCA method as used by Mann et al. effectively mines a data set for hockey stick patterns. Even from meaningless random data (red noise), it nearly always produces a hockey stick.”

    That was the message at the heart of well-executed PR campaign.

  9. Lars Karlsson says:

    “…whom I nicknamed Chewbacca [16] for less obvious reasons, … ”

    Chewbacca defense?

  10. jsam says:

    The Auditor’s delicious lack of self-awareness is an asset. It is what makes him such a formidable ClimateBollocks player.

  11. I have yet to commit what some, but not me, may consider the most important fact.

    Will do so later.

    Will also explain how I came with Brandon’s nickname, and why I don’t use it that way anymore.

    ***

    If you like Berne, you may also like what I call the Scientist Game, M2:

    http://neverendingaudit.tumblr.com/post/13157516404

    Games people play was my theme in the comment thread at Keith’s:

    http://blogs.discovermagazine.com/collideascape/2011/11/18/games-people-play/

    Keith’s title was too good to miss.

  12. Willard says:

    Reblogged this on ClimateBall ™ and commented:

    A guest post on another Climate blog.

  13. miker613 says:

    Tried to follow this post, but couldn’t. The part I’m familiar with – McIntyre’s recent quote “MM05 did not imply that Mannian PCs “typically” produced hockeysticks: it stated it.” – was perfectly clear to me, and seems to be completely accurate. [My paraphrase: “In summary of this post, I want to say that MM05 was not merely “implying” that Mannian PCs usually produce hockeysticks. On the contrary, it was a clearly stated result of the paper that hockeysticks are in fact _almost always_ produced – unlike what some are presently claiming.”] You seem to have decided to make yourself crazy over the statement, slicing and dicing it upside down. I don’t want to be harsh, but McIntyre is not responsible for your reading comprehension.

    McIntyre has another post today, again trying to get across his point about hockeysticks. I think those who want to argue with him may want to go over there, as Nick Stokes is, trying to defend their counter-claims. No one but a Climateballer should be continuing to quote “you only get one hockeystick in a hundred” stuff when McIntyre has published a rebuttal, at least until it has been refuted.

  14. miker613,
    The point, which is not that complex, is that what I said was referring to how people interpret the paper, not what the paper said specifically. I find it hard to see why that is hard to understand. The point I was getting at, though, was that people interpret MM05 as implying/suggesting/stating that if you took 70 random timeseries, none of which have a trend, that the MBH98 method will produce a hockey stick. That, I believe, is not correct and should be obviously not correct, since if the 1902-1980 average is that same as the 1400-1980 average, for all of the 70 timeseries, then the centering can make no difference. What MM05 shows is that if you have 70 timeseries, each of which has a lot of persistence (i.e., decadal variability), then the MBH98 method will regularly produce a result with a HSI that is non-zero (either positive or negative), while the MM05 method will produce a range that include zero (i.e., the distribution functions in Figure 2 that Steve shows in his most recent post).

    So, as I see it, if the average trend of your 70 timeseries is flat, the MBH98 method can produce a non-zero HSI, while the MM05 method does not. However, this does not mean that each of the 70 timeseries is flat, simply that the average of all 70 is flat. Given that – in reality – each timeseries is sampling the same overall climate, having a sample of 70 timeseries, each of which is highly variable but in which the average is zero, does not appear particularly representative of reality.

  15. miker613,

    No one but a Climateballer should be continuing to quote “you only get one hockeystick in a hundred” stuff when McIntyre has published a rebuttal, at least until it has been refuted.

    Ahh, that’s not what people are saying. People are pointing out that most figures that have ever been presented from the MM05b analysis are from a sample of 100 chosen on the basis of having the 100 highest positive HSI values. Hence, suggesting that this sample is typical is not correct.

  16. > You seem to have decided to make yourself crazy over the statement, slicing and dicing it upside down.

    That rings a bell:

    Willard, have you made efforts there to get your comments fixed? Or do you prefer paranoia? RC moderation is not some theory; they boast about it. I have been going to climateaudit for an awfully long time, seen lots of critical comments and very little moderation except for Off Topic.

    http://judithcurry.com/2014/09/11/fraudulent-hockey-stick/#comment-628298

    ***

    > I don’t want to be harsh, but McIntyre is not responsible for your reading comprehension.

    That rings a bell too:

    Don’t mean to be insulting, but anyone who is defending Cook’s study after Duarte’s articles just doesn’t care about anything but winning.

    http://judithcurry.com/2014/09/05/what-exactly-is-going-on-in-their-heads/#comment-625045

    The title of Judy’s post echoes the attribution in MikeR’s first quote.

    ***

    The statement in MM05b was a bit stronger than “typically”, MikeR.

    Find it and report.

  17. miker613 says:

    ATTP, there is not much point trying to convince me: I’m not really in a position to follow the math, at least not without taking time I don’t have. Most of my post wasn’t directed at the math point, which is above my pay grade, but at Willard’s confusion about what McIntyre was claiming.

    As for the math point, I can only repeat: McIntyre has posted a response, which sounded convincing (to a person like myself, anyhow, whatever that means). Why leave Nick Stokes on his own over there? Most people who trust him are there, not here; why not convince them? I saw Robert Way do an excellent job over there defending his paper, and I thought that the clear result was that people over there respect him and his work – at least most of them.

  18. Joshua says:

    ==> ” and I thought that the clear result was that people over there respect him and his work ”

    Interesting.

  19. miker613 says:

    “People are pointing out that most figures that have ever been presented from the MM05 analysis are from a sample of 100 chosen on the basis of having the 100 highest positive HSI values. Hence, suggesting that this sample is typical is not correct.” If I follow you, you’re saying that he over-egged his paper, showing atypically good results. He claims that the results are nevertheless correct: over 95% of results show more than one s.d. hockeysticks. So does that mean you’re accepting his point as well? (I don’t know if he has conceded or responded to yours.) Because that is not at all what I’m seeing. The impression that I’m getting is that loads of people out there think that their favorite blog(s) are saying that M&M’s result was cherry-picking; i.e., they are defending Mann. Are you accepting that M&M are right that Mann’s methodology was dead wrong?

  20. > Willard’s confusion about what McIntyre was claiming.

    Show me, miker613.

    Since you claim not having understood the post, I’m not sure you’re in a position to claim this.

    ***

    > McIntyre has posted a response

    A response to what, where, and how does this matter for the price of tea?

    If you’re here to peddle in the Auditor’s editorials, at least do it properly.

    ***

    > I saw Robert Way do an excellent job over there defending his paper, and I thought that the clear result was that people over there respect him and his work – at least most of them.

    That rings a bell:

    Willard, have you made efforts there to get your comments fixed? Or do you prefer paranoia? RC moderation is not some theory; they boast about it. I have been going to climateaudit for an awfully long time, seen lots of critical comments and very little moderation except for Off Topic. See Nick’s response. I saw Robert Way do an excellent job of defending his paper there, and others.

    Op. cit.

    I have no intention of playing the spam game anymore, Miker613. When the Auditor will implement login via Twitter accounts, I’ll reconsider. Perhaps I’ll enquire when I’ll have more time.

    He’s welcome to comment here. And at Nick’s, where he never goes. Think about it, miker613: you’re asking me to do something the Auditor never does regarding Nick.

    Thank you for your concerns.

  21. miker613 says:

    Joshua, don’t you think so? I thought Way was brave, and effective, and Steve Mosher backed him up too, both at climateaudit and at judithcurry. There are of course still fools at both places; I haven’t noticed a shortage of them at any website I’ve ever seen.
    Another good example is Pekka. I think he has done a great job of keeping some of us honest at judithcurry. Doesn’t mean he is always right, but he is always smart, always sensible and always courteous. Why _wouldn’t_ that be more effective than sneering?
    I can’t understand why both sides build their fortresses, crouch inside them, and preach to their choirs. What do they expect to accomplish?

  22. miker613 says:

    “Show me, miker613.” Willard, I did. That was the paraphrase in brackets. Not sure how you understand McIntyre’s words, but that is what he meant.

  23. Joshua says:

    ==> “Joshua, don’t you think so? ”

    I look at Rud’s comments in a recent thread – directed at Way and accusing him of deliberate fraud (he wrote a book, doncha know. A book with chapters)* – and a near total lack of blowback from the SWIRLCAREs that reside there. That doesn’t seem at all inconsistent with the treatment of Cowtan and Way that I see at Judith’s more generally (I don’t read CA enough to comment – and I acknowledge that your original comment was describing that specific context)…

    So the absence of evidence (defense of Way against accusations of fraud) in that case is not evidence of disrespect for Way, but within the larger context (including comments about Way and/or his work that I would be willing to bet could not fairly be characterized as respectful) I do think it is entirely consistent with the pattern of fortress-building and choir-preaching that you describe.

    ==> “I can’t understand why…”

    Really, Mike? You really can’t understand it?

    It is so ubiquitous, in fact, so characteristic not only in the climate wars but in myriad similar political proxy battlefields. I think it is fairly simple to understand.

    * Sorry, couldn’t resist.

  24. miker613,
    One reason I haven’t gone over to Steve’s is that I’m trying, and failing, to have a break. Plus, I do have less patience than Robert Way and Nick Stokes. I don’t think that there is a simple answer to this question

    Are you accepting that M&M are right that Mann’s methodology was dead wrong?

    My broader answer is that I’ve actually found the MM05 quite interesting and have learned a lot while trying to improve my understanding of it.

    The impression that I’m getting is that loads of people out there think that their favorite blog(s) are saying that M&M’s result was cherry-picking; i.e., they are defending Mann.

    I think the issue is that most figures that anyone has seen from the MM05b analysis look a lot like MBH98 hockey sticks. However, these are not randomly selected from all 10000 simulations, they were selected on the basis of having the 100 highest positive hockey stick index (HSI). What I think people also don’t realise is that this hockey stick shape is not a consequence of the method, it’s because the data being analysed (the 70 randomly generated timeseries) actually has a hockey-stick-like trend in it. The AR1(0.9) model used by MM05 to generate their random dataset has a large persistence, which really just means that a future data point depends on past points. This means that if one data point has a high value, the next one has a good chance of also being high. The 0.9 means that this can continue for quite some time, hence one can generate a hockey stick like shape randomly.

    The more interesting figure in MM05b is their Figure 2, which shows the difference between the MM05b (centered) and MBH98 (decentered). This shows that the MM05b has a smooth distribution of HSI with a peak at zero and extending symmetrically to positive and negative values. The MBH98 method produces two peaks – one positive and one negative. This indicates that the MBH98 method can produce non-zero HSI values even when the MM05 method produces values close to zero. One can certainly interpet this as the MBH98 method producing a hockey stick when there isn’t one in the data. I’m not convinced that this is quite correct given that the underlying timeseries are randomly generated with a lot of persistence. Hence a HSI of zero using MM05b probably means that the average trend is zero, not that all 70 timeseries have zero trend. In this case the centering may well produce a different answer but how realistic are 70 completely random timeseries, all of which are highly variable but in which the variability averages to a zero trend.

  25. > Not sure how you understand McIntyre’s words, but that is what he meant.

    That does not show any confusion from my part. Whatever the Auditor really meant, miker613, his demonstration clearly misrepresents what Anders was saying. He just took the same words Brandon furiously kept recursing and editorialized on it.

    Please note: [discussing] what Anders said there fails to address Nick’s main point. This was the main fact I had yet to commit. The Auditor simply [indulged in] a red herring.

    The red herring worked by invoking a strawman: the “ClimateBaller doctrine.” This misrepresents the concept of ClimateBall ™. This misrepresentation is itself a ClimateBall ™ move.

    This was the main point of this post.

    You are not addressing it, miker613. You’re running with (paraphrasing) “but what the Auditor really means,” “but the Auditor’s new post,” “but Science” (whatever this means) and “I dare you to tell that at Steve’s”.

    ***

    So as I see it, you’re basically reproducing the Auditor’s move by not addressing the main point of this post. You claim incomprehension, so you move to what you claim understanding, which misrepresents what I said. Then you hammer that the Auditor has a new post, whence this has nothing with this [very] post, [which, i.e. the Auditor’s post] addresses Nick’s main point in part only, and quite indirectly if you ask me.

    Since you ask that we concentrate on the math stuff, that you claim not understanding the math stuff, and keep promoting math stuff, I suggest you check out the meaning [of] “doctrine,” miker613.

  26. miker613 says:

    Oh, well. No idea what you are trying to do, Willard; I freely admit it. But I think it is more important to understand what McIntyre is actually saying than to shoehorn him into your picture of things. Given one of Anders’ comments here, perhaps McIntyre misunderstood him. But that doesn’t seem to be an option you’re dealing with.

  27. Look, miker613.

    The Auditor conflated what “people” believe, say some proposition P, with the fact that people believe P, which was the object of Ander’s claim. Refuting P does not change the fact that people believed P at the time the statement was made. Anders’ claim remains true, and his “untrue” uppercut fizzles.

    In the case of the Auditor’s editorial, the P in question had nothing to do with Nick’s main claim. I don’t think it has anything to do with any of Nick’s claims, for that matter. The Auditor used “ClimateBaller doctrine” in a magnificent guilt trip by association.

    Is that clearer?

  28. JCH says:

    Steve Mosher on Robert Way – pal review?

  29. anoilman says:

    miker613: There is no value in understanding McIntyre.

    I looked at his web site, and he spent some 6 months complaining about how crappy he felt the peer review process was. 6 months, complaining that his word was more valuable than all others. 6 months complaining that they should listen him and only him.

    He’s got delusions of grandeur, and I don’t think I need to see anymore.

    Sometimes when you submit a paper for review, you get one live wire. The odd ball, crack pot. The guy out there, that thinks all in the world needs to be re-stylized to suit his vision. The only thing new here is that McIntyre blogged about doing it.

  30. uknowispeaksense says:

    The auditor, as you like to call him, really isn’t worth your time. While he does say…..stuff…..that stuff is meaningless. If he really had anything worthwhile to contribute to the faux debate, he would demonstrate his alleged statistical prowess and publish something ….somewhere …….anywhere that isn’t the dodgiest of dodgy E&E. Even though he does seem to be a lot smarter than Anthony Watts, I actually rate him lower because he doesn’t use his skills to produce anything meaningful. I would even go as far as sating that if Steve McIntyre is bagging you, you are on the right track. Just beware his flying monkeys. You could end up in my position if one of them goes psycho. Now its time for my tenuously related link.

    http://uknowispeaksense.wordpress.com/2012/09/11/steve-mcintyre-butthurt-about-comment-moderation-however/

  31. miker613 says:

    http://climateaudit.org/2014/09/28/t-statistics-and-the-hockey-stick-index/#comment-733106
    Posted Anders’ comment (https://andthentheresphysics.wordpress.com/2014/09/29/the-ghost-of-present-climateball-tm/#comment-32633) at climateaudit, got a reply. We’ll see.
    “Steve: when he says “most figures” are based on high-HSI values, I presume that he means Wegman Figure 4.4, a figure that was produced long after our articles had received considerable publicity and which attracted negligible contemporary attention. The relevant MM05 figure, Figure 2, is based on all 10,000 simulations. I haven’t gotten to a discussion of Wegman Figure 4.4 yet, as I wanted to first clear up issues about orientation and the “hockey stick index”, which ClimateBallers use to move the pea, but I do plan to discuss it.”

  32. Eli Rabett says:

    Eli had tried to put a couple of comments over at the Auditor’s lair, but alas they have not appeared although they were nice and calm. The Auditor (Eli may call him that, may he not?) is taking this very seriously.

    To reiterate, by defining the Hockey Stick Index as McIntyre did and putting software out there that selected the 100 with the highest scores for the random draw, those pointing downwards at the end (which would have negative HSI) would not even be considered. Thus this whole nonsense about flipping the birds is just that.

    Perhaps someotherbunnies could go post there. Eli tires of agitating a bag of wind.

  33. miker613 says:

    “The AR1(0.9) model used by MM05 to generate their random dataset has a large persistence, which really just means that a future data point depends on past points. This means that if one data point has a high value, the next one has a good chance of also being high.” That sounds interesting. Wouldn’t real data from tree rings (or real thermometer data) be like that too? Whatever variables drive it, surely it has persistence from year to year as well, not just totally jumping around. It sounds right – am I misunderstanding it? Can’t someone tell what the general statistics of tree ring data are?

    But that doesn’t mean it would have hockey sticks in Mann’s sense, does it? Some would go up over here and down in the middle and back, some down over here… Presumably a few would be mostly lower on the left and higher on the right. Now you could average everything and get pretty much no trend, but Mann’s method picks out the hockeysticky ones that have an appreciable difference between the mean and the values at the end. Am I understanding it correctly?

  34. uknowiss,
    Yes, I retweeted your relevant post about your stalker. I worry that I might have one of my own.

    miker613,

    The relevant MM05 figure, Figure 2, is based on all 10,000 simulations.

    Yes, I realise and I didn’t say otherwise. I also expressed my thoughts about Figure 2 in my earlier comment. My point (and I think the point that Nick Stokes has been making) is that most (if not all) hockey stick figures from the MM05b analysis (whether in MM05b itself, or in the Wegman report) came from the sample of 100 that were selected on the basis of having the 100 highest postive HSI values.

    The issue with this (and is what lead to my comment that Steve then included in his post) is that people conclude that the MBH98 analysis always produces a hockey stick like feature that looks like the MBH98 result. Given that all that most have seen is a hockey-stick-like result from MM05b that happens to be on of the top 1% in terms of HSI, this is not obviously the correct interpretation.

    As I said in my earlier comment, I do think that Figure 2 is the more interesting figure, but am unconvinced that randomly generated timeseries with high persistence is the correct comparison with MBH98. For example, I do not think that the MBH98 method would produce a result with a high (positive or negative) HSI if all 70 timeseries had no trend. I think the reason that the MBH98 method produces a two peaked HSI profile is because what it is analysing is randomly generated timeseries with high persistence (variability).

    FWIW, I can’t work out the plural of timeseries.

  35. miker613,

    But that doesn’t mean it would have hockey sticks in Mann’s sense, does it? Some would go up over here and down in the middle and back, some down over here… Presumably a few would be mostly lower on the left and higher on the right. Now you could average everything and get pretty much no trend, but Mann’s method picks out the hockeysticky ones that have an appreciable difference between the mean and the values at the end. Am I understanding it correctly?

    Okay, yes it seems that if you have a set of highly variable timeseries but with an average trend of zero, the MBH98 method produces a result with a non-zero HSI (either positive or negative). My point was that this is not necessarily representive of a set of actual tree-ring data (for example). Also, if you had a set of timeseries that weren’t variable, each of which had a zero trend, then I don’t think that MBH98 analysis would produce a result that had a large (positive or negative) HSI. Given that the tree-ring data is meant to be sampling the same overall climate, why would we expect each timeseries to be highly variable, but uncorrelated?

  36. John Mashey says:

    In the real world, a useful skill set is that of assessing expertise, something the Dunning-Kruger-afflicted especially lack.

    Perhaps people would assess the credibility and knowledge about this of folks like Grace Wahba and Noel Cressie, in comparison with McIntyre or his supporters at CA.

  37. Michael 2 says:

    “FWIW, I can’t work out the plural of timeseries.”

    It’s incomplete — two adjectives and no noun. More complete (and whose plural is then obvious) is “a time series record”, or “we have several time series records”.

  38. MikeN says:

    ATTP, yes decentered PCA as Mann did it will not produce hockey sticks from random data with no trend. It mines for hockey sticks, elevating them to the top PC. Also, if there is a proxy with higher Medieval Warm Period than modern, it could flip that.

    There is a second step after the PCA, that also filters for hockey sticks.

    The end result of all this is that of the 70 proxies, only 22 make it into the hockey stick.
    http://hiizuru.wordpress.com/2014/02/18/manns-screw-up-3-statistics-is-scary/
    Robustness is clearly an issue.

  39. Michael 2 says:

    John Mashey says “a useful skill set is that of assessing expertise, something the Dunning-Kruger-afflicted especially lack.”

    Nice demonstration of the Dunning-Kruger phenomenon 🙂

    http://en.wikipedia.org/wiki/Dunning%E2%80%93Kruger_effect

    People are not “afflicted” with Dunning-Kruger. It is a word that describes both expert and inexpert persons; experts tend to deprecate their own expertise while unskilled persons tend to exaggerate their expertise.

    The reason is simple — the more you learn the more you realize the magnitude of what *could* be learned but has not yet been learned.

    A person just starting out in computer programming creates “hello, world!” and feels creative and powerful. That is good, because it is a long road ahead of him or her.

    Your demonstration reveals that you know the words “Dunning-Kruger” but clearly do not understand the meaning thereof or you would not be carelessly calling it an “affliction”.

  40. I’ve been faced with the denier mantra of “Mann’s methods produce hockey sticks even with random data” coming from “skeptics” for years now. This has been their killer blow that “proves” his work was wrong. No quibbles, no caveats. Just point blank, “Random noise produces hockey sticks.” This is what the majority of “skeptics” argue.

    It seems to me that it’s rather pointless how McI manages to explain this away. If he thinks the interpretation coming from “skeptics” is wrong then he should make a clear and definitive statement of that fact, rather than trying to wrestle the nuances.

  41. anoilman says:

    I just love this video!
    “Steve MacIntyre going berserk!”
    “He’s hitting everything in sight!”
    “The linesmen are trying to control MacIntyre!”
    “He’s gone Bananas!”

    miker613: If you are confused about the Willard and Climateball, you should read the article on it.
    https://andthentheresphysics.wordpress.com/2014/03/26/climateballtm/

    Willard is in fact more concerned with how people are talking than with content. I believe he is biased towards to the consensus view of climate change. However he is pretty agnostic about who he goes after and why. He doesn’t care for trolling from anyone.

    The hard reality is that those bloggers/deniers out there are training a lot of bad behavior not conducive to real technical discussion. In a formal discussion there is a lot of back and forth. (Sometimes with violent exclamation points.) However its hard to understand a technical point when your opponent blindfolds the referee, and goes for an illegal tackle. Perhaps a sucker punch to the nuts when no one is looking.

    Willard loves to point that kind of behavior out to people, like so;
    https://andthentheresphysics.wordpress.com/2014/09/29/the-ghost-of-present-climateball-tm/#comment-32639

  42. anoilman says:

    John Mashey: “In the real world, a useful skill set is that of assessing expertise, something the Dunning-Kruger-afflicted especially lack.”

    Steve McIntyre is a business executive with no expertise. Case closed.
    http://www.desmogblog.com/steve-mcintyre

  43. miker613 says:

    “Given that the tree-ring data is meant to be sampling the same overall climate, why would we expect each timeseries to be highly variable, but uncorrelated?” Dunno. But is it or isn’t it? I don’t _expect_ it, but I can easily imagine that (a) tree rings widths are driven by numerous effects that last for years (changes in surface temperatures, changes in humidity, changes in surrounding plant life, nearby forest fires, deer migration,…), but that (b) surface temperature, while relevant, is dominated by some of the other confounding variables which are very local. You would end up with AR1(0.9) [or something] with very little temperature trend. Mann’s method will find a hockey stick anyhow. Again, I’m checking – am I understanding this correctly?

    “The issue with this (and is what lead to my comment that Steve then included in his post) is that people conclude that the MBH98 analysis always produces a hockey stick like feature that looks like the MBH98 result. Given that all that most have seen is a hockey-stick-like result from MM05b that happens to be on of the top 1% in terms of HSI, this is not obviously the correct interpretation.”
    As I posted before, McIntyre has said pretty clearly that this _is_ the correct interpretation: that is, you will (almost) always get a hockey stick of at least one s.d. Your comment “Given…” doesn’t deny that, it just claims that he picked better examples than he should’ve, higher s.d., not that the more normal examples wouldn’t also prove his point.

    As for your comment “Given… top 1%…”, it seems to me that the McIntyre comment I posted claims that this is not true. He claims that he did no such thing, and that Wegman’s Fig 4.4 is something that he will yet discuss. So I will have to wait for that piece of the issue. Are you sure that McIntyre did cherry-picking himself?

  44. KR says:

    John Mashey – Indeed, assessing expertise is rather critical to evaluating claims.

    In the case of McIntyre and MM05, there is a considerable lack of expertise demonstrated by _not_ evaluating their PCs for significance, as MBH98 did, and said lack leads to their dropping the significant PCs with the 20th century ‘hockey-stick’ (discussed here). The MM centering convention changed the number of PCs from MBH’s, redistributing the basis functions/coordinate systems that describe the principal components of the data. Evaluating those PCs for significance is an undergraduate level task at which they utterly failed.

    Oh, and their ‘red-noise’ model is absurd, given that they didn’t detrend the proxies first (leaving hockey-stickiness in the ‘noise’), and having utterly absurd persistence in their noise, leading to non-stationary series. Fail.

  45. > I wanted to first clear up issues about orientation and the “hockey stick index”, which ClimateBallers use to move the pea […]

    [And then] Auditor blames unidentified opponents of what he himself is doing with his posts about orientation. His latest post is not about orientation anyway. A fierce ClimateBall ™ player we have there.

    Don’t you see what the Auditor’s doing, miker613?

  46. miker613 says:

    “It seems to me that it’s rather pointless how McI manages to explain this away. If he thinks the interpretation coming from “skeptics” is wrong…”
    I’m still in Bizarro World here. He says they’re right, he says it very clearly, he repeats it. [Are you maybe focused on the AR1(0.9) issue that ATTP is discussing? My point would still stand, assuming he thinks the trees are similar. Otherwise, I have no clue what you mean.]

  47. > I’m still in Bizarro World here.

    That rings a bell:

    This is getting stranger and stranger.

    http://neverendingaudit.tumblr.com/post/87101071969

  48. MikeN says:

    Random noise will produce hockey sticks if you have enough of it. Mann uses correlation weighting, screening by correlating to temperature, and decentered PCA, all of which will elevate hockey sticks in the data. If you have enough random data, then there will be some hockey sticks in there.

  49. anoilman says:

    Why does anyone care about ‘Auditor’ complaining about an old outdated paper? More importantly, why can’t he put on his big girl panties and explain how all the other methods, datasets, etc are wrong? Seriously, the hockey stick has been proven repeatedly using other methods and data sets.

    My personal favorite was isotopes in water extracted from underground by oil companies in Alberta. Yup… hockey stick. I suppose that means that oil companies are now behind this ‘conspiracy’…

    The simple explanation of why he has to absolutely dwell on Mann, and wear blinders to all else, is that he simply can’t disprove it. He doesn’t have the hutzpah to do a real or thorough technical job. He has to focus on the confusion he’s spun around Mann.

  50. KR says:

    AOM – Actually, it’s (IMO) because it’s far more convenient in rhetorical terms to demonize and denigrate a single person, using them as a (inappropriate) stand-in for the science, than to actually present data that might support McIntyre’s point. If, of course, there actually _were_ such data, which seems not to be the case.

    McIntyre is a one-trick pony. And it’s a lousy trick…

  51. miker613 says:

    anoilman, who do you expect to convince? Those of us who have been following climateaudit for a while know better. He posts on all those other methods and data sets as well. Most of them share some of the same issues. Other issues have been laid to rest; Mann’s use of PC was one (till recently). As Robert Way said in 2010, “you should warn those doing that particular one [the “Climate’s changed before” post at SkS] to stay away from Mann’s 2008 paper if they take this topic as it seems it has actually been invalidated by climate audit”.
    McIntyre is posting on it now because he saw others reopening the topic. On many other topics, there are interesting discussions on his blog, with experts from other sides chiming in and sometimes making good points. Sometimes his friends make good points too.
    So what am I going to think of you? Someone open-minded, or a total partisan cheerleader?
    Look, you have “proof” that McIntyre is no expert. Your proof? A post on a blog you like. Fine, you like that blog. But why would that convince anyone who doesn’t?

  52. MikeN,
    Okay, thanks. Maybe we can clarify something then. Here are the 3 scenarios I can envisage.

    1. There is no trend in any of the 70 timeseries. In this case the MBH98 method will not find a hockey stick.

    2. There is a net trend in the 70 timeseries. In this case the MBH98 method will find a HS as will the MM05b method, MBH98 might amplify this slightly.

    3. The 70 timeseries are highly variable, but the average trend is zero. MBH98 finds a hockey stick when MM05b does not.

    Is that about right? If so, then if our millenial temperature history is highly variable with a zero net trend, maybe MBH98 would be wrong. Since it appears not to be, this seems less of an issue than some might suggest.

    There are still two things that I’m uncertain of. In case 3 above, does the result depend on the number of retained PCs? Also, why did MM05b need to use the MBH98 data to generate the random red noise with persistence?

    An additional comment I was going to make earlier was that – IMO – the discussion today of MBH98 and MM05b is largely irrelevant (at least scientifically, which is all the interests me). Our broad understanding today is not significantly different to what it was 16 years ago. Our climate has been roughly stable for about 1000 years with some variability (MWA, LIA) and that we’ve warmed by about 0.9 degrees since the mid-1800s and, today, are probably/likely warmer than we’ve been for at least 1000 years. That, in a sense, is all that really matters in terms of understanding our climate. All the rest just seems like – as Willard is I think trying to illustrate – is just ClimateballTM.

  53. MikeN,

    If you have enough random data, then there will be some hockey sticks in there.

    Sure, but that’s because they’re actually there. The real data isn’t random. I hope you’re not doing what I’ve started to call a Keenan.

  54. anoilman says:

    miker613: There are quite literally thousands of papers on the hockey stick. McIntyre hasn’t scratched the surface.

    You think thousands of phds got it wrong and an industry executive got it all right… hmm… OK. Good luck with that. (I think the UK is allowing barbers to do surgery again as well, that’s probably sufficiently skilled… for you.)

    I’d direct you to what KR said, “it’s far more convenient in rhetorical terms to demonize and denigrate a single person, using them as a stand-in for the science, than to actually present data that might support McIntyre’s point”.

  55. KR says:

    MikeN – Principal components can be considered basis functions, coordinate systems, for expressing the variation in a data set. Given different data patterns, not all coordinate systems are equally expressive. Consider Cartesian XYZ versus polar RθΦ coordinate when describing an arc on a sphere – you need fewer equations to express that arc in polar coordinates than XYZ’s.

    MBH and their centering of data on the instrumental/calibration period produces a PC coordinate system that quite concisely expresses the recent hockey stick, and requires only two (2) PCs to encompass most of the proxy records variation. MM05 and their centering is less efficient, requiring five (5) PCs to _equally_ express the variations in the records – and the hockey-stick pattern is there in PC4. Which would have been included had they actually performed the tests MBH did.

    Yet McIntyre and McKitrick utterly failed to test their PCs for significance, retaining only two as per MBH, leaving out much of the record varations – an undergraduate level error. Therefore I (and others) conclude that they are _not_ actually experts. Just loud.

  56. KR,
    A question for you then. Is the difference in the histograms shown in Figure 2 of MM05b largely because they only kept 2 PCs when using the centered method. In other words, if they had kept all 5 PCs when using the centered method would that histogram have matched the histogram representing the MBH98 method?

  57. > An additional comment I was going to make earlier was that – IMO – the discussion today of MBH98 and MM05b is largely irrelevant [.]

    I disagree. This thread is becoming a textbook example of a food fight about irrelevancies. Please, do continue. So far, miker315 is earning the first star:

    https://en.wikipedia.org/wiki/Three_stars_%28ice_hockey%29

    May the Ghost of ClimateBall ™ Yet to Come relish this moment!

  58. I disagree. This thread is becoming a textbook example of a food fight about irrelevancies.

    Isn’t that what I was saying? 🙂

  59. Sometimes, irrelevancies can be quite relevant, AT.

    Look! A squirrel that looks like a hockey stick!

  60. Joshua says:

    Food fight about irrelevancies?

    In the climate wars?

    No way!

  61. anoilman says:

    If one is going to jump sharks one must first obtain shark repellant;

  62. Are you suggesting that the Auditor baited AT with a red herring in a blog post about something Nick Stokes hasn’t said, Oil Man?

  63. Eli Rabett says:

    KR raises an interesting point. The best data is the instrumental record from ~1880, the blade. If the proxys do not correlate (or anticorrelate) to the instrumental records they are not useful as temperature proxys.

    Since we know the instrumental record on a regional and local scale since 1880, we can evaluate proxy usefulness on an individual basis by seeing how well they agree with the instrumental record (this is what the entire dendro food fight about the lack of agreement since ~1960 is about).

    Agreement with the instrumental record since 1880 is a valid method for deciding whether proxys should be included in a reconstruction of global and regional temperature. As it turns out for trees, the best proxys are for trees which are growing in marginal environments (ask Hughes) such as the Sierra or the Gaspe.

  64. anoilman says:

    No. I’m flinging food.

  65. KR says:

    ATTP – Fig. 2 in MM05 is showing the distribution of just PC1 as rated by the MM HSI. Since the centering convention MM used shifts the hockey-stick from PC1 to PC4, this is wholly unsurprising. And again, demonstrates that McIntyre and McKitrick should have take a class or two in principal component analysis. Fig. 2 is the result of looking in the wrong box…

    As shown by Mann at RealClimate, and by the published analysis in Wahl and Ammann 2007, when appropriate significance selection is used, the results of centered/decentered PCA are almost identical.

    They won’t be exact matches, as they are different coordinate systems for expressing the same variations, and they won’t have exactly the same sum significance level or dimensional coverage – but if you include 5 PCs in MM05 as per the MBH selection rules, as shown in Fig. 1 of the RealClimate page, MBH PC 1-2 and MM05 PC 1-5 show the same results.

  66. anoilman says:

    If McIntyre wrote about shark jumping this is probably what he’d write. This is funny folks, give it a read. Its not like we need to take the real McIntyre seriously.
    http://www.newyorker.com/magazine/2014/07/28/deniers-2

  67. KR,
    Thanks, I did wonder about that. And I’ve just gone and reread the figure caption which clearly says PC1. So, I think I’ve been a bit generous in some of my earlier comments. It’s not that the MBH98 method can find hockey sticks when the MM05b method does not. It’s that the MBH98 method can find hockey sticks in PC1 when MM05b does not.

  68. John Mashey says:

    Sigh. Nobody (at least not those supporting McIntyre) wanted to comment on Grace Wahba and Noel Cressie. I wonder why that was ,but they could lean from Strange Scholarship… which mentions both a few times, see especially p.20, 53, 54, 56-60, 67, 134, and especially Cressie’s comments to Wegman, too late for the report, but Wegman never backed off. What Wahba told Wegman isn’t public, but it was ignored, too p.42-43.

    Hint: look them up in Google Scholar. Look up Edward Wegman or Steve McIntyre.
    Can any tentative assessments be drawn? (by other than D-K -afflicted)

  69. KR says:

    It’s my impression that McIntyre and McKitrick got most of their statistical ‘expertise’ from reading the R language help files, rather than a more encompassing study of fundamentals. Their failure to understand the fundamentals shows here with their failure to check PC significance, with McIntyre’s attempts to (mis)understand exploratory factor analysis WRT Dr. Lewandowsky’s work by again naively throwing R functions at it (Dr. Oberauer, here), and with McKitrick’s failure to convert degrees to radians in four years of economic modeling.

    Not. Experts.

  70. KR,
    Indeed. It would seem that Ross McKitrick still hasn’t improved his understanding basic statistics.

  71. KR says:

    ATTP – It’s not that MBH can find PC1 hockey sticks and MM05 cannot. It’s that the hockey-sticks are in the data, and due to MM05 throwing away three significant PC’s they dropped it from their analysis.

  72. miker613 says:

    “Agreement with the instrumental record since 1880 is a valid method for deciding whether proxys should be included in a reconstruction of global and regional temperature.”
    No. Unacceptable. That is guaranteed to produce a hockey stick. Start with our AR1(0.9) and hundreds of proxies. Screen out all that don’t match recent data. What’s left will average flat in the past (since it is AR1), and for some reason all go up since 1880.
    You need to take some subset of proxies, see which match the temperatures since 1880, use that to try and make a hypothesis about proxies and temperatures (maybe something like “trees that are growing in marginal environments”). Then you go back to the rest of your proxies and take all that fit your hypothesis, whether or not they went up. Check if they match the instrumental record since 1880. If they did, use their data from the past as well. If they didn’t, start over. You may find that your proxies just don’t work well.
    You aren’t allowed to use the data that you used to form your hypothesis for validation.

  73. KR,
    Indeed, that’s what I’ve been getting at in some of my earlier comments, but I worded it poorly there.

    miker613,

    That is guaranteed to produce a hockey stick. Start with our AR1(0.9) and hundreds of proxies. Screen out all that don’t match recent data. What’s left will average flat in the past (since it is AR1), and for some reason all go up since 1880.

    I don’t get this. We’re talking about real proxies here and simply suggesting that they should match what is known. That doesn’t guarantee they’ll be flat prior to 1880.

  74. miker613 said… “You aren’t allowed to use the data that you used to form your hypothesis for validation.”

    Hm, since when was the instrumental record a “hypothesis?”

  75. KR says:

    MikeN – If a potential proxy does _not_ replicate temperatures independently confirmed by instrumental means, it’s not a good proxy. And that has to be accompanied by good physical reasons _why_ a proxy might reflect temperatures, such as marginal growth zones where temperature has an effect, otherwise I could for example start making claims about orbital patterns based on proxy evidence of the relative numbers of black and grey squirrels.

    AR1(0.9) has absolutely nothing to do with proxy evaluation, and the rest of your comment is equally nonsense.

  76. Try this, KR:

    http://climateaudit.org/econometric-references/

    I would thread this path lightly if I were you.

  77. I don’t think that there has ever been any dispute on the validity of the decentred method of MBH98 if all time series used contain a clear blade caused by the recent warming. Thus there should not be any dispute if all time series have been correctly validated before they are included in the set of time series to be used.

    The above leaves open the questions, whether all time series satisfy the above criteria, and whether screening based on validation can be done without introducing a bias by the screening.

    Under somewhat weaker conditions the method may still work, but gradually the problems build up and the risk that variability in the earlier periods is erroneously lost in the analysis grows.

  78. ligne says:

    “otherwise I could for example start making claims about orbital patterns based on proxy evidence of the relative numbers of black and grey squirrels. ”

    wait, isn’t that the gist of Scafetta’s latest paper?

  79. miker613 says:

    “It’s that the hockey-sticks are in the data, and due to MM05 throwing away three significant PC’s they dropped it from their analysis.” It still seems to me that everyone here is making the same mistake that Eli is making. If you have a lot of persistent variability in the data, there will be totally random, totally spurious hockey sticks in it. If you then use a statistical analysis that selects for them, you will find them. Eli suggested one such way: keep everything that went up since 1880. Mann had another. Everyone here seems to be defending these things, and attacking the messenger who pointed out that this isn’t allowed, and I can’t understand it.
    This is why we insist on scientists preparing their statistical analysis before they start gathering data. Otherwise it is no problem coming up with analyses that “find” what you want.

  80. KR says:

    Pekka – “all series” is an overstatement, as there will always be proxies in a large set that do not reflect past changes exactly the same. Rather, the PCA method is used to determine the primary variations, the minimal set of dimensions, that can express the variability in _all_ the series put together.

    As to validation, there are now many reconstructions using various and sundry proxies and statistical methods (NOAA reconstructions, Air Temperature, Global and Hemispheric), which broadly agree. The consilience of results indicates that the majority of proxies are in agreement, which in my opinion settles the question of sufficient validation. Gradually the evidence builds up, and the risk of error diminishes.

    I _cannot_ see how proxies can be chosen without some validation screening.

  81. > It still seems to me that everyone here is making the same mistake that Eli is making.

    Does that include me, miker613? I don’t recall having opined on this. Please refresh my memory.

    ***

    Also, you seem quite proficient at all this out of a sudden:

    I’m not really in a position to follow the math, at least not without taking time I don’t have.

    https://andthentheresphysics.wordpress.com/2014/09/29/the-ghost-of-present-climateball-tm/#comment-32640

    I thought your point was about my “confusion,” which you seem to have dropped now that you have something else to talk about in this thread.

  82. KR says:

    MikeN – If you use PCA (properly), the proxy data indicates a hockey-stick. If you don’t use PCA, the proxy data indicates a hockey-stick. If you look at other proxies entirely, you find hockey-sticks. Like the spaghetti sauce commercials – it’s in there. The recent anomalous rise in temperatures is _not_ a statistical artifact.

    The hockey sticks in the MM05 (erroneously named) red noise runs are due to unreasonable persistence choices, to including hockey-sticks in the generation of the ‘noise’, and to a 100:1 sorted subset selection. As such they really have little relevance to evaluating the proxy data.

  83. miker613 says:

    “And that has to be accompanied by good physical reasons _why_ a proxy might reflect temperatures, such as marginal growth zones where temperature has an effect,”
    ‘Hm, since when was the instrumental record a “hypothesis?” ‘ It’s not. The hypothesis is that this type of trees is an accurate proxy.
    “I don’t get this. We’re talking about real proxies here and simply suggesting that they should match what is known. That doesn’t guarantee they’ll be flat prior to 1880.”

    All these comments have the same issue. We don’t _know_ what is a good proxy. “Good physical reasons” offers a possibility, a hypothesis. It still has to be validated. It may be wrong, because other confounding factors may overwhelm the temperature dependence.
    If you form your hypothesis and lots of the trees fail to match since 1880, you can’t just pick the trees that succeed. What has happened is that your hypothesis was wrong. You need a new, more accurate hypothesis, that (more effectively) includes the ones that succeeded and excludes the ones that failed.
    Then you _need more data_. That is so you can validate your new hypothesis. We are never allowed to validate a hypothesis on the training data, which is the data you used up already.
    (One way to do this properly is to hold out some of the data originally when you were training.)

    So no, Mann is not allowed to look for the statistical method that is “best at finding hockey sticks”. That’s data snooping, and can be expected to give spurious results.

  84. anoilman says:

    KR: “I _cannot_ see how proxies can be chosen without some validation screening.”

    All data needs to be cleaned and screened. In the case of tree rings, it needs to be detrended, or its wrong.

  85. miker613 says:

    Willard, I gave up on understanding what you were talking about. So no, you were not included.

    And thanks for the compliment, but I am not very proficient. I’m a programmer, not a statistician. I’ve taken a couple of courses in statistical data analysis, enough to react when people say things that are clearly wrong. What I’m saying is the basics of how to do any scientific study where data analysis will be used.

  86. Joseph says:

    I have heard others say that proxy analysis has progressed since 1998 and that even if Mann’s methods were not perfect, it does not mean it wasn’t a valid paper and didn’t make a significant contribution to a new field.

  87. KR says:

    AOM – Indeed, part of _developing_ a proxy is to identify and remove influences not related to your variable of interest. The detrending of growth rings is primarily to correct for differing growth rates in young/old trees; there is far more material in outer rings than inner, and the detrended residual is the useful proxy.

  88. miker613 says:

    “The hockey sticks in the MM05 (erroneously named) red noise runs are due to unreasonable persistence choices, to including hockey-sticks in the generation of the ‘noise’, and to a 100:1 sorted subset selection.” I hear you. I also hear them: McIntyre has very clearly said exactly the opposite in his recent posts. Not being a statistician, I’m not in a position to judge, and I’m not at all interested in the ad hominems that some of you think are relevant.
    I have seen various sources from outside the skeptical blogospher that I found trustworthy that agreed with McIntyre; I mentioned one already. If I list more I imagine you will find ad hominem attacks on them as well, so I won’t.
    I will repeat the point I made earlier: if you care about convincing people like me, addressing McIntyre and co directly on his blog is the way to do it. They are currently discussing this very subject. Posting links to blogs you like isn’t.

  89. > Willard, I gave up on understanding what you were talking about.

    I was talking about the Auditor’s trick to turn a discussion about what Nick Stokes does not want you to know, a misleading claim if you ask me, into a food fight. He proceeded this feat by misrepresenting both Anders (AT) and the concept of ClimateBall ™. He still does as we speak [regarding ClimateBall ™ at least, if we consider your quote of him in this very thread, above].

    Your involvement in the thread shows how such baiting works, and your contributions offer a good example of what he may have meant by a doctrine.

    Is that clear enough, this time?

    Please, don’t lose hope. It’s always possible to communicate.

    ***

    > So no, you were not included.

    Who was, then?

    Quotes might be nice.
    Thanks.

  90. KR says:

    MikeM – “Mann is not allowed to look for the statistical method that is “best at finding hockey sticks”. That’s data snooping…”

    They didn’t. They (not just Mann, also Bradley and Hughes, that’s an ad hominem targeting on your part) applied machine learning techniques for identifying and expressing primary variations in complex data (PCA) to temperature proxies, and that’s what came out of analyzing the data. You are accusing MBH of scientific fraud, which is not only insulting to them but utterly disproved by the _many_ reconstructions that followed. On many blogs you would be snipped for that alone.

    Enough. You are clearly neither familiar with the math nor listening to any critical discussion of MM05.

  91. anoilman says:

    Interesting;
    “If you form your hypothesis and lots of the trees fail to match since 1880, you can’t just pick the trees that succeed. What has happened is that your hypothesis was wrong. You need a new, more accurate hypothesis, that (more effectively) includes the ones that succeeded and excludes the ones that failed.”

    It is also foolish not to screen and clean data. Multiple papers have been written by real dendropaleoclimatologists on the suspect trees. Why? It looked like there was something wrong.

    Anyways, Briffa gave up on all this, and produced a data set with all the trees in it. (The problem trees aren’t as significant since there are now way more trees in the data set.)
    http://www.cru.uea.ac.uk/cru/people/briffa/yamal2009/

    What does Steve say on all this? Crack Cocaine he says. Nice. Real professional. A standup kind of guy.
    http://climateaudit.org/2009/09/26/briffas-yamal-crack-cocaine-for-paleoclimatologists/

  92. miker613 said… “If you form your hypothesis and lots of the trees fail to match since 1880, you can’t just pick the trees that succeed.”

    Isn’t that kind of the point, mike? To identify which tree ring series respond to temperature? In order to do that you have to validate and calibrate which trees actually respond to temperature. Dendro folks spend a huge amount of time and effort doing this validation and calibration work. You can’t just build a statistical process and then throw all the tree ring data in the blender and expect to see a rational result.

  93. anoilman says:

    KR: Thanks for that. Most of the papers I found on tree rings talked a lot about rain fall. (All the papers I read also confirmed regional hockey sticks.)

  94. KR says:

    AOM – Any low frequency (slow) variation unrelated to your variable of interest can add a spurious trend, including regional rainfall. My personal preference with growth-ring analysis (trees, teeth, otoliths) would be to cross-validate against independent proxies (speleothems?) to check for drift, and to use sufficiently spread tree proxies to cancel out smaller regional effects.

    Sufficiently broad (spatially) tree growth-ring proxy sets do just that – regional effects cancel out.

  95. anoilman says:

    KR, Rob Honeycutt: That is correct. The idea with a proxy temperature set is to first validate the proxy with actual temperature measurements, then extrapolate backwards.

    In the early Yamal tree ring set, many trees didn’t correlate (temperatures trended down), so they were discarded. Briffa has since put those trees back in the data set with a lot more trees that have been harvested. Case closed.

    This concept isn’t much different from geospacial temperature cleaning.

  96. KR,

    In all statistical analyses even remotely similar to this case the requirements of guaranteed objectivity (not introducing bias) and power are contradictory by nature. One must make some compromise that satisfies well enough both requirements. When there’s a lot of data of sufficient quality the compromise is likely to be safe, but when there’s a shortage of data and when noise level is high no compromise is fully satisfactory.

    It’s natural that first multi-proxy analyses of paleoclimatic time series were done in a situation where the volume of high quality data was rather low. It’s right to do the first analyses at that time, but the resulting limitations must then be fully recognized.

  97. anoilman says:

    KR: How do you know if you have sufficient temperature coverage? i.e. global

  98. KR says:

    Pekka – I would largely agree; MBH89 was a first attempt at using those data reduction techniques, a seminal paper in proxy reconstructions. Later reconstructions including more proxies, ongoing proxy methodology development, and improved statistical techniques are certainly more accurate.

    But that doesn’t make MBH wrong, just limited by the data and statistics of the time. And it certainly doesn’t justify ongoing McIntyre tirades accusing them of incompetence or fraud, particularly when MM05 has such glaring errors.

  99. Marco says:

    “if you care about convincing people like me, addressing McIntyre and co directly on his blog is the way to do it”

    I don’t think we can ever convince you, until McIntyre admits he was wrong or did something wrong. Which he is unlikely to do, since his whole reputation depends on having such a wonderful criticism of MBH98/99. Just imagine he says “yes, I made several boo-boos when criticizing MBH98/99”. Seriously imagine it. I can’t.

    In other words, whatever any of us writes on climateaudit, it will be dismissed one way or another by moving goalposts or other forms of deflection. Just ask Nick Stokes about that. Or take the case when people found out McIntyre had the Yamal data for *years* already, provided to him by the actual data owner. McIntyre’s response to that revelation? “I wasn’t sure it was the same data Briffa used”. Maybe I missed it, but I don’t think any of the regulars at climateaudit expressed any disappointment he had had that data for years and yet kept on making insinuations about problems with the data without doing any analysis, because he claimed he didn’t have that data.

  100. anoilman says:

    Pekka, that is my understanding as well. Early work was bound to have concerns around coverage, and quality. Hence, more aggressive data cleaning as opposed to now.

    Initially Mann thought they didn’t have enough data of sufficient quality. It was only on further analysis that he thought they could derive something of statistical significance.

  101. Marco says:

    Pekka, I urge you to read the title of MBH99.

  102. KR says:

    AOM – Ideally you would sample with sufficient density to cover the globe via observed temperature anomaly correlations (Hansen and Lebedeff 1987), and extrapolate as needed via current temperature correlations where you cannot (mid-Pacific rather lacking in pine trees).

    Lacking that or even with the best data you make the best estimates possible, hopefully with an estimate of potential errors due to undersampling and proxy uncertainties (such ranges are one of the major contributions of MBH98), and limit your results appropriately – which is why there are numerous Northern Hemisphere reconstructions (including MBH) and fewer Southern Hemisphere and global ones.

  103. Just a personal recollection from the years I was active in physics research.

    I learned soon that the most successful of my colleagues added always in their papers mild overstatements of the importance and topicality of the work. That helped greatly in getting the papers published, but it was also important that the overstatements could not be taken as outright and intentional falsehoods.

    I would expect that the decades since have not changed the above, and that the approach was not restricted to theoretical elementary particle physics.

  104. Pekka,

    It’s right to do the first analyses at that time, but the resulting limitations must then be fully recognized.

    In a sense, yes. However, this happens by newer studies identifying better techniques or problems with the original technique. It doesn’t require that the original authors publicly announce the limitations in their first study. It becomes obvious to anyone who works in the field. As others have mentioned, this hounding of Michael Mann is extremely unsavoury (and that is a polite way of putting it).

  105. > [T]his hounding of Michael Mann is extremely unsavoury […]

    Nobody expects the Auditor’s inquisition, AT.

    The Climateballers’ Doctrine Makes Him Do It

  106. Pekka,

    I learned soon that the most successful of my colleagues added always in their papers mild overstatements of the importance and topicality of the work…..it was also important that the overstatements could not be taken as outright and intentional falsehoods.

    Exactly, knowing how to make your work sound novel and interesting without saying anything that is untrue, can be a valuable skill; one that I don’t think I actually have. To be clear, this can be an issue when people make their research sound much more interesting than it warrants. Knowing precisely where to draw the line is, however, not always completely obvious. Most who regularly cross onto the wrong side, get a reputation for doing so.

  107. KR says:

    Pekka – Upon reflection I must say that I find your most recent post more than a bit concerning. You seem to be insinuating that MBH was overstated by the authors, and therefore may be less worthy than currently discussed, without coming out and clearly saying so – I don’t consider that justified. As I recall Mann et al were quite surprised by the attention paid to their paper, and much of the ongoing discussion by Dr. Mann is in response to repeated and unjustified criticism, not self-promotion. Your comment only makes sense if the authors promotion of their own work was somehow anomalous – and I really don’t see that as the case.

    A far better case might be made for papers like Spencer and Braswell 2011, which expressed uncertainties (“an unsolved problem”) while their press release states “Earth’s atmosphere is apparently more efficient at releasing energy to space than models used to forecast climate change have been programmed to “believe.””. Now _that_ is overstatement.

  108. ATTP,
    Actually every paper should be explicit of its limitations, if they are not fully obvious to every scientist even without an explicit statement. That’s a strong requirement, but the practices of writing papers and review procedures are unfortunately not always on that level.

  109. KR,
    I can’t really speak for Pekka, but I just thought he meant that criticising people for making their research sound interesting is silly, as that is what most do. I didn’t take it to be aimed at MBH98 specifically or at all. If I understand Pekka, I tend to agree. Noone’s encouraged to write a boring abstract and if you can’t make your research sound interesting, what made you do it in the first place?

  110. Pekka,

    Actually every paper should be explicit of its limitations

    Sure, but only if you know them beforehand. If you’re doing groundbreaking work, others may later find better ways to do the analysis. That doesn’t mean that the original authors should then publicly denounce their first attempt. I wasn’t referring to people who knew the limitations but didn’t mention them. I was referring to situations where the limitations may only become apparent once more research has been done and new techniques have been developed.

  111. Eli Rabett says:

    Pekka, there is an interesting Email from Malcolm Hughes to Mike Mann somewhere in the old MBH 98 archive which says something like, here are some more proxys that better fit the needs of the paper (SM is well aware of this). Eli takes this to mean that all possible tree ring series may not be useful proxys for local temperature and precipitation because some other known factors were limiting.

    So yes, there was a selection of proxys that showed significant change in line with the instrumental record. If BH had chosen potential proxys that did not change during the instrumental period, they would have simply contributed noise to the reconstruction.

  112. > Actually every paper should be explicit of its limitations, if they are not fully obvious to every scientist even without an explicit statement.

    And by magic, this returns us to Nick’s claim:

    Steve McIntyre [the Auditor] has a new post up at ClimateAudit. It’s called “What Nick Stokes won’t show you”. It’s a continuation of the smokescreen about demanding the unselected PC1s be shown with orientation favorable to a hockey-stick interpretation, using a hockey stick index (HSI), rather than as his program produces them. Again pretending that it’s about Wegman aligning the orientation, rather than selecting the top 1% by HSI without disclosure.

    http://moyhu.blogspot.com/2014/09/more-climateball-at-climate-audit.html

    My emphasis.

  113. Joshua says:

    Anders –

    This is quite a break you’re taking,. 🙂

  114. Joshua,
    Yes, not really doing a good job of staying away 😉

  115. Joshua says:

    ==> “Yes, not really doing a good job of staying away 😉 ”

    I’d just blame Willard. Seems to work for Mosher and Stevie-Mac.

  116. KR says:

    ==> “Yes, not really doing a good job of staying away 😉 ”

    It’s the squirrels.

  117. KR,
    I try to keep my comments on the level of principles and avoid accusing specific people. Unfortunately I cannot always avoid referring to some particular case as the comment might be very difficult to understand otherwise. Here I referred to first multi-proxy analyses, because that seemed to make the point more precise.

    I haven’t rechecked recently how each of the papers describes the uncertainties. Thus I don’t make any claims about that.

  118. I’m not the one who went to talk to Brandon, Joshua:

    This may not go well, but I’ll give it a try

    http://hiizuru.wordpress.com/2013/12/12/basic-truths/#comment-15

    The ghost of present ClimateBall ™ goes beyond the Auditor’s misrepresentation of it.

  119. Yes, but at least my initial assumption turned out to be correct 🙂

  120. I declare one area that I don’t know enough about and that seems be overlooked often. That’s the significance of spatial variables in the reconstructions. It’s relatively easy to understand how PCMs appear in a single time series, but including the spatial variables makes the issue more difficult to handle intuitively.

    MBH98 shows in Figure 2 five empirical orthogonal functions that reflect the spatial effects in the analysis.

  121. Eli Rabett says:

    Pekka, the spatial reconstructions are also available in principle for each year. One of the other figures shows a couple of years.

  122. KR says:

    Pekka – “Each of these eigenvectors is associated with a characteristic spatial pattern or ’empirical orthogonal function’ (EOF) and its characteristic evolution in time or ‘principal component’ (PC).”

    Those five EOF’s are what are displayed in Fig. 2, not PC’s – linear combinations of those simpler patterns account for most of the spatial patterns in the data, much as you can combine strong spatial or temporal frequencies in Fourier space to roughly reproduce data showing those strong frequencies.

  123. AT,

    Just imagine what this ClimateBall ™ food fight would have been if your comment thread had many entry points, like Judy’s.

    Also, I suggest that you follow some kind of business hours, at least tonight, since Rachel will be MIA for a while. You could close down this thread for the night, or send everyone in the pending bin. I have not moderated this thread to date, and I won’t moderate it.

    You can always reopen tomorrow morning, if you like how this goes. I still have notes too, but must get going.

    Thank you for your patience,

    w

  124. Pekka… The MBH papers were actually quite modest in their conclusions. If you haven’t read them, I believe now would be an appropriate time. As well, they provided very broad uncertainty ranges on their graphs. I’m not sure what more you would have expected.

  125. KR says:

    Note that spatial reconstruction skill from the proxy network is demonstrated against instrumental data, clearly shown in Fig. 4.

  126. Eli,
    It’s unlikely that I’ll end up publishing anything any more. Thus I have the luxury of having my main interest in improving on my conceptual understanding. Even before I have valued highly the role of intuitive understanding of the physics – including everything related to QM, where new concepts must be learned.

    Statistics is another area where reaching correct intuitive understanding is often both rather difficult and potentially important. Adding dimensions to the system being considered may mean that a lot of time should be spent on analyzing the actual data before correct intuitive understanding is reached. I’m not sure, how far I’m ready go go on that, and even less sure, whether the early analyses are the right place to spend the effort.

  127. Rob,

    I have read them, but not very recently, at least not in full.

    The argumentation on the correctness of papers published 15 years ago is really out of place, but the technical discussion that the papers raise is still intriguing, and it’s sometimes difficult to avoid making some comments.

  128. Willard has a point and I need some sleep, so I’ll turn on moderation and look at things in the morning.

    KR, that Figure 4 in MBH98 is very impressive.

  129. KR,
    Results like those shown in the Figure 4 belong to those that I feel that I don’t know enough. It may be highly significant, but I have the suspicion that perhaps it’s not after all, and instead somehow forced by the method. I don’t write this to criticize the paper but rather to tell, how I look at any paper of this nature (meaning a paper that applies new methods that have not been scrutinized by independent scientists).

  130. miker613 says:

    “Again pretending that it’s about Wegman aligning the orientation, rather than selecting the top 1% by HSI without disclosure.” Seems to me that McIntyre was perfectly clear about this. Kevin O’Neill mentioned both issues as examples of McIntyre’s fraud. McIntyre said he would start with the easier/smaller one first, so he did, with a post or two showing that aligning the orientation was a non-issue, and has been done by everyone till now except Nick Stokes.
    Stokes and O’Neill responded with numerous claims that that wasn’t the issue, picking the top 1% was the issue. Which is fine, it’s a concession on the smaller issue (which was indeed something that O’Neill had mentioned, I saw it). Now McIntyre has to finish dealing with the larger issue. What’s your problem with all this, aside from you don’t like McIntyre?

  131. miker613 says:

    “They (not just Mann, also Bradley and Hughes, that’s an ad hominem targeting on your part).” Goodness. I was being brief.

    “You are accusing MBH of scientific fraud, which is not only insulting to them but utterly disproved by the _many_ reconstructions that followed.” (a) I am not accusing him of anything of the sort. Sloppy data analysis is a common problem, and using a unusual statistical method is not fraud, even if it turns out to be wrong. Data snooping is an easy trap to fall into, even with the best of intentions. (b) Using an unacceptable method of analysing data is not in any way disproved by more successful reconstructions afterwards.

  132. JDS says:

    I do have one question about this debate and I’d be interested in hearing from anyone from either side of the debate on what may be a very naive question or something already discussed (sorry if this is the case). I think I understand the statistical issues reasonably well, but in the midst of all this debate I’m not sure what the point of doing reconstruction with proxies that only go to 1980. The 1880-1980 blade from what I understand is not due to CO2 (except a few years in the late 1970s). So what does MBH99 say – that temperatures were relatively flat for a 1000 years, then suddenly go up from 1880-1980 for unknown reasons? Yeah, the extended (human-caused) blade from the instrumental record (1980-) looks scary, but it looks like it is piggybacking on an already non-human caused blade from 1880 to the late seventies

    It seems to me the real hockey stick debate is not so much over PCA methods or even about AGW. We already know warming from late seventies onward is human caused. So the issue is why were temperatures already high before AGW began in the late seventies? It seems there are two implications based on what type of reconstruction you believe:
    1. MBH99 and other reconstructions using Mann’s PC1 – the 1880-1980 non-human portion of blade is unprecedented over a 1000 years. If this is the case, we need to do a lot of research to find out what natural factors caused such an unprecedented spike prior to the advent of AGW in the 1980s and 1990s. 1880-1980 non-human caused “blade” is a huge mystery if these reconstructions are correct.
    2. Ljungqvist 2010, Esper 2012, etc. – natural factors have caused similar spikes before in MWP and Roman Warm Period. 1880-1980 spike no big deal. AGW in 1980s and 1990s comes at a bad time on top of a natural spike similar to MWP and RWP. In this case 1880-1980 blade is not such a mystery, just part of some natural cycle that has been going on for the last 2000 years.

    Again, I’m just a lurker so sorry to ATTP and others if this is a naive question or it has been discussed before.

  133. miker613 says:

    “Or take the case when people found out McIntyre had the Yamal data for *years* already, provided to him by the actual data owner. McIntyre’s response to that revelation? “I wasn’t sure it was the same data Briffa used”. Maybe I missed it, but I don’t think any of the regulars at climateaudit expressed any disappointment he had had that data for years and yet kept on making insinuations about problems with the data without doing any analysis, because he claimed he didn’t have that data.” In other words, you don’t like McIntyre and will always assume he’s lying. They do like McIntyre and will assume he is telling the truth. What is surprising about that? Why couldn’t he have been telling the truth?

  134. KR says:

    JDS – It might be better to discuss more recent data and analysis, such as list under NOAA reconstructions, Air Temperature, Global and Hemispheric. And note that these include many reconstructions not using PCA techniques.

    Proxies are quite useful when instrumental data is unavailable. But I see _no_ reason not to include the highly accurate and detailed instrumental data that includes the last 30 years – with mercury expansion or thermocouple conductivity as the temperature proxies.

  135. Eli Rabett says:

    JDS at the time of MBH 98, there were a lot of proxys that went to 1980 and many fewer that went to 1998. Since the method required a uniform cut off date 1980 was a reasonable choice which kept a large number of proxys, and even then some of the proxys had to be padded a few years.

    On this subject every mole needs whacking.

    This was a subject of, shall Eli say, contention back in USENET days.

  136. anoilman says:

    JDS… here’s a graph from Ljungqvist 2010… it confirms the seriousness of AGW. We are already way beyond natural forcing.
    http://www.skepticalscience.com/ljungqvist-broke-the-hockey-stick.htm

    I also believe AGW started before 1970. Its not a sudden thing.

  137. John Mashey says:

    JDS: watch Bill Ruddiman’s Tyndall Lecture at AGU in 2013 and then take a look at the book.
    That’s very strong evidence that Earth’s climate hasn’t been entire;ly natural snce humans started agriculture, and neither Roman nor MWP were entirely natural. See also 2000 years of CO2 from Law Dome.
    Since the Holocene started:
    1: 100% natural
    2: enough people and heavy-footprint agriculture to modify the normal interglacial, specifically, retard the normal slow cooling
    3: ~1AD- enough people that one can see the jiggles in CO2 and CH4.
    That includes the Roman period, the dip afterwards, the rise to the MWP (what there was of it), and the stronger dip into LIA (50Mperson die-off in Americas + volcanoes + Maunder). Then
    Without humans, CO2 would have been down around 240-250ppm about time the next started.
    4: ~1800-1850 CO2 heads up fast and never looks back. That’s fossil fuels + (maybe) some more deforestration. At this point, we’ve totally overpowered the natural curve.

    Most of the Holocene is *not* entirely natural and why people argue over how much of it should be renamed the Anthropocene.

  138. Chic Bowdrie says:

    KR at 7:27 pm: What is the difference between data snooping and applying “machine learning techniques for identifying and expressing primary variations in complex data (PCA) to temperature proxies?”

    Also, which actual temperature measurements are used to validate proxy temperature sets? Is it the global mean or actual thermometers somewhere close to the trees?

    Re Attp’s reply to Pekka at 8:09 pm: Forgiving Mann’s statistical streches as “mild overstatements of the importance” of a “first study” is one thing. Using it as the centerpiece of an IPCC report is quite another.

    A point that I think miker613 and MikeN made and, which I don’t think was clearly refuted, is that MBH98 weighted certain proxies relative to others based on a criteria that they had a hockey stick shape. If this is not true, there should be definitive evidence showing no weighting was involved. Anyone know where?

  139. verytallguy says:

    Chic

    If this is not true, there should be definitive evidence showing no weighting was involved. Anyone know where?

    And then there’s Mann’s wifebeating Chic.

    Where’s the definitive evidence that ever stopped? 

  140. verytallguy says:

    Mann and pages2k

    Definitive. Evidence.

  141. verytallguy says:

    Willard,

    I think the Auditor is indeed a fierce player.  Here’s you and everyone else at ATTPs talking about Mannian fraud. 

    Well played. 

  142. anoilman says:

    Chic Bowdrie: Calibrating sensors is standard practice Chic and Mikes… There’s nothing to refute here.

    I believe the detrended (Yamal) tree rings were compared to temperature data to ensure accuracy for the time periods for which there is overlapping ring/temperature data. This is not the whole ‘stick’. (Obviously we wouldn’t use proxies if we had direct temperature measurements.)

    Tree rings which diverged from temperatures for which we have measurements were pruned from the data set, as being ‘suspect’. There are many papers on those specific tree rings written by actual experts. (Who are not named McIntyre.) I think they looked at sulpher contamination and a few other things.

    This is pretty much how all sensors are calibrated. For instance you can see on this data sheet that we have a simple quadratic linearity error which can be corrected for in software. In can be detrended with a generic curve, or individually calibrated for accuracy. (Most folks prefer RTDs for accuracy, I grabbed this inaccurate sensor for an example.)
    https://www.sparkfun.com/datasheets/DevTools/LilyPad/MCP9700.pdf

    If you want hockey sticks in your tree ring data set, a sure bet is to not detrend that data. I believe McIntyre did not detrend his data sets.

  143. The case of the MBH98 paper is a good example of difficulties that we meet in discussion related to climate science. The paper presents a good case for discussing methods of statistical analysis and to learn from such discussion, but it has become virtually impossible to participate in the discussion without getting involved in personality wars.

    I for one think that think that McI has presented, and continues to present points that are at least worth taking seriously, but I don’t want to condemn in any way the authors of the original paper for writing and publishing the paper. Discussing those points of McI or the technical choices done in the original work seems still interesting and relevant for learning to understand better all statistical analyses of multiple proxy records (including those that do not use PCA as many problems are more general).

    A concrete case helps greatly in focusing the discussion. Therefore this case would be so useful input for the discussion, but it seems that a discussion of the type I would like to see is impossible on a open blog.

    Similar problems apply equally to discussing GCM type climate models. Technical discussion of some details is certainly possible as it’s separate enough from more general implications, but discussion of more general nature gets difficult.

    For the case of the multi-proxy analysis an approach similar to that of Science of Doom is probably possible. That means starting from basics and building gradually towards understanding full analyses and their limitations. That requires only someone competent enough in both understanding the issues and in writing in a clear and interesting fashion – and willing to take the effort. In this case the papers of MBH would represent only steps on the way towards the present understanding. That might help in circumventing much of the personality dispute.

    Large GCM type models are much more complex, if they were not SoD had probably already covered them as they fit otherwise well his agenda.

  144. JDS,

    The 1880-1980 blade from what I understand is not due to CO2 (except a few years in the late 1970s).

    I’m not sure where you’re getting this from. The latest evidence suggests that most of the warming since 1950 was anthropogenic. Even this has to be interpreted carefully. It doesn’t mean that it was probably 51% anthropogenic. The best estimate is that anthro provided more than 100% of the warming since 1950 (i.e., other influences produced some cooling). Also, it doesn’t mean that anthro didn’t contribute prior to 1950. If you look at the changes in forcing since the mid-1800s, the biggest is anthropogenic. This really means that the reason we’re warmer today is almost all because of the changes in anthropogenic forcing. This doesn’t, however, mean that all warming in all time interval was anthropogenic. There have been periods were solar played bigger and smaller roles, volcanoes were either more or less active, and internal variability probably played a role on decadal timescales. The underlying trend of the blade of the hockey stick is almost certainly predominantly anthropogenic.

    Chic,

    Forgiving Mann’s statistical streches as “mild overstatements of the importance” of a “first study” is one thing. Using it as the centerpiece of an IPCC report is quite another.

    Firstly, are we talking papers or the use of papers? Secondly, is what the MBH98 reconstruction showed wrong? There’s nothing yet to suggest that it is, so I see no reason why it shouldn’t have been used by the IPCC as it was. Also, see my response to VTG below for more.

    VTG,

    I think the Auditor is indeed a fierce player. Here’s you and everyone else at ATTPs talking about Mannian fraud.

    Except that I’ve found this quite useful. I was willing to consider that Figure 2 in MM05b illustrated that short-centering produced hockey sticks when standard centered didn’t (given persistence in the data). As KR has pointed out, though, this Figure only shows PC1. His suggestion is that short-centering requires 2 PCs while standard centered requires 5. This RealClimate post appears to say the same and suggests that the short centering produces a signal in PC1 while standard centered does so in PC4. If correct, what would be very interesting is a comparison between a HSI histogram for PC1 when short centering is used, and one for PC4 when standard centering is used. If KR and the RealClimate post are correct, then these two histograms should be about the same and what is being illustrated in MM05b is largely irrelevant and simply a consequence of not considering all the relevant PCs.

    You may, therefore, be making a valid point that I have been too critical of MBH98 and not sufficiently critical of MM05b.

  145. verytallguy says:

    ATTP,

    fair enough, it is a physics blog, so I guess a deep understanding of the stats in a seminal paper from 16 years ago is valuable.

    from an engineer’s perspective however

    -no sceptic has ever produced a reconstruction showing anything other than a hockeystick (despite constant whining about the “disappearance” of theapparently obvious MWP)
    -multiple other reconstructions have showed broadly similar results
    -regardless as to what they show, reconstructions of the last two millennia are only a minor part of the consilience of evidence that CO2 drives climate
    -the blade of the stick is confirmed by instrumental record anyway
    -on the shaft, if anything, reappearing the MWP through different methodology would imply a significantly *higher* sensitivity(!)

    and from a PRs perspective
    -the more people debate fraud the worse light climate science is put in, regardless as to the balance or conclusion of any such discussion.

  146. VTG,
    I agree with your engineer’s perspective completely.

    I also agree with your PR view. I do worry that this just rehashes old things that we should no longer be discussing and it can be remarkably childish. The interesting thing, though, is that it seems that the 1:100 cherry-pick is now out in the open and hard to dispute. Also, Steve is now focusing on Figure 2 of his paper which shows that the PC1 using short-centering can produce a non-zero HSI even when standard centering does not. If KR and RealClimate are correct, then this is simply a consequence of not comparing the correct PCs and if someone (I’ve left a comment at Nick Stokes’s about this) can produce a histogram of the standard centering PC4 and if that looks the same as the short centered PC1, then Steve will have to argue why his comparison of PC1 was the correct comparison. It might be an interesting outcome (although I will add that JeanS’s response to Nick Stokes on Steve’s latest post doesn’t bode well).

  147. The controversy is strongly linked to the existence of two different framings of climate change.

    The one that I and most of the followers of this site, I’m sure, consider correct is that the potential risks of AGW were foreseen based on understanding of the atmosphere and Keeling’s observations in the 1970s, i.e., before any clear signs were present in the temperature record, and that the understanding has since then been confirmed by further research. (Personally I learned about that in 1980 from the IIASA activity that produced the book Energy in a Finite World).

    The alternative framing dismisses all the above and claims that nothing else matters than direct observations. In this framing the question of the temperature of the MWP and internal variability get overemphasized. Therefore these people consider the shape of the shaft of the hockey stick essential. They try to show that we cannot exclude even strong and rather fast variability, including perhaps periods with warming as fast as the recent one.

    For us who accept the first framing it would certainly be valuable to know much better the details of climate of past centuries, but those are not essential for final conclusions. They might be, if the data would be of really high quality, but whether the knowledge is poor or still worse is almost irrelevant.

    To me the details of multi-proxy analyses are very interesting from intellectual perspective, not as a source of information essential for climate policy decisions.

  148. Where did I discuss what you think I am discussing, Very Tall?

    Most of my comments were dedicated to miker613, who plays like a ClimateBall star player out of sudden, after a slow start where he was merely raising concerns. He still does raise the concerns, about my mind state for instance, something for which I am thankful. I’ll return to these concerns in a moment.

    His move “I don’t understand you” to switch to the Auditor’s bait was a thing of beauty, don’t you think? It improves upon Brandon’s “you make no sense!” quite a lot. He did play some suboptimal moves in his latest comment he addressed to me, though. He alluded to Nick for the sake of making a physical play on me, and will soon have to be reminded that his peddling has no bearing on anything Nick claims.

    ClimateBall ™ takes discipline.

  149. ” It improves upon Brandon’s “you make no sense!” quite a lot. “

    A reminder of how absurd the situation is when a principal mover and shaker of skeptical thought, Brandon, is the Chewbacca defense personified.

    Count the number of times that BS has used “you make no sense!”, and then other howlers such as comparing climate modeling to video games:

    “I’ve created several (relatively simple) models when discussing strategies in video games that had perfect precision and accuracy.”

    This guy has to be a prank.

  150. > I do worry that this just rehashes old things that we should no longer be discussing and it can be remarkably childish.

    I doubt anyone but ClimateBall players are reading this, AT. Even I skim most of the hockey stuff. Too much words. Not only here, but on every other ClimateBall fields, including the Auditor’s.

    If hockey stick matters are so complicated after so long, the Auditor deserves some part of responsibility. His bitching does not help clarify matters. His identity politics obscure his verbiage:

    http://chronicle.com/article/Why-Academics-Writing-Stinks/148989/

    That the Auditor pulls you into this was suboptimal, more so considering the way he did.

    If he wanted to portray himself as not being a ClimateBaller, he failed.

  151. Eli Rabett says:

    To repeat (well Mike is the 613th Mike) Steve Mc flipped everyone the bird when he changed HIS definition of the hockey stick index. HIS code pushes negative going curves to the bottom of the rank order which HIS code then sorts on to find HIS 100 curves with the most positive hockey stick index from which HIS code then selects 10. HE DONE IT.

  152. Kevin O’Neill and Nick Stokes got their hands dirty and did everyone a great service when they dove into the McIntyre mudpit.

    Like I have said before, the hardest thing to do is to debunk an analytical mess. One wonders why Mann didn’t do this himself, but then you realize that you would be stooping down to the level of an ankle-biter.

  153. KR says:

    WHT – Actually, Mann has responded a number of times, as per later publications (Mann, Zhang, Hughes 2008), as well as multiple discussions on RealClimate.

    As far as I can see McIntyre isn’t really presenting anything that hasn’t already been debunked by Mann or by Wahl and Ammann 2007, which states:

    “When proxy PCs are employed, neither the time period used to “center” the data before PC calculation nor the way the PC calculations are performed significantly affects the results, as long as the full extent of the climate information actually in the proxy data is represented by the PC time series.” (emphasis added) – which MM05 failed to do.

    MM05’s primary error is that they failed to estimate PC significance, and thus dropped three PCs. including PC4, the one with the strong ‘hockey-stick’ signal. Everything past that is, well, “…a tale … full of sound and fury, Signifying nothing.”

  154. And, Steve can’t seem to avoid making some kind of suggestion of fraud, saying in his latest post

    By (perhaps inadvertently) drawing attention to Mann’s 2005 “Preisendorfer” calculation, ATTP has raised an issue that is highly relevant to Steyn’s allegations of fraud and which ought to be minutely examined by Steyn and others.

    I also have no idea how I’ve been dragged into this as I don’t remember specifically drawing attention to Mann’s 2005 Preisendorfer calculation. ClimateballTM?

  155. KR says:

    I’ll add another serious omission in the MM05 work – the lack of validation statistics. Wahl and Ammann 2007 compute these for both MBH and MM, finding that _all_ of the MM suggested changes (dropping proxies, using too few PCs, etc) that result in significantly different reconstructions fail validation over the differing periods.

    Assertions without validation, as in MM05a/b, are never a good idea…

  156. > ClimateBall ™?

    Pretty much. You Made The Auditor Do It, after all. (That’s in Berne’s book, M2.) You say things about hockey sticks.

    I too, perhaps. That the Auditor fumbled on his first Get Your Facts Straight tentative never prevented him to follow through his tackling. On the contrary, since he’s attempting some kind of reverse takeover on the concept of ClimateBall:

    Northwest Exploration Company Limited carried out gold exploration as a provate company in the mid-1990s. In late 1997, it was taken over by Northwest Explorations Inc. through a share exchange to become a 100% subsidiary of Northwest Explorations Inc. and continued to carry out gold exploration as a subsidiary of Northwest Explorations Inc. I was President of the subsidiary and a director of the parent. Northwest Explorations Inc. went public and continued gold exploration. The gold exploration was unsuccessful and all gold exploration operations ceased and all employees of the subsidiary were terminated. In 1998, CGX Resources did a reverse takeover of Northwest Explorations Inc. and changed the name of Northwest Explorations Inc. to CGX Energy Inc. As a result of the takeover, I ceased to be a director of the parent company. CGX Energy Inc. had no interest in the gold exploration subsidiary, Northwest Exploration Company Limited, which, at the time, had some gold exploration concessions of negligible value, all of which have since been dropped, and some small liabilities and sold the subsidiary to a private group with which I was involved. Northwest Exploration Company Limited has successfully done some property transactions involving hardrock mineral exploration properties (nothing to do with oil and gas) acquiring some assets in the process.

    http://climateaudit.org/2005/07/24/crowley-and-north-1991/

    This kind of paragraph makes me think that academic writing emerges from contractual practices.

    The Auditor is a fierce ClimateBall ™ player.

    ***

    The Get Your Fact Straight is an important part of his gas lighting business.

    http://neverendingaudit.tumblr.com/post/78452403963

    This episode was also about Steyn.

    As Kevin might wonder, what would be the odds of that?

  157. Chic Bowdrie says:

    Verytallguy,

    Thanks for the response. However, the pages2k plot does not answer my question, because I would still need to examine MBH98 AND pages2k to see if both use the same methodology.

    Anoilman,

    “Calibrating sensors is standard practice ….”

    Thanks for addressing that issue. Not being a statistics expert, it’s hard to know what is valid and what is not. The way I interpreted the point that Mike613 made is that mining the data for trees that correlate with temperature assumes the hypothesis that tree ring data is a good proxy for temperature. If only trees that correlate well with temperature over a defined time interval are used for calibration, it still leaves the possibility that some other factors confounded the correlation. But that is not really my main concern. I want to know if the methodology Mann used favored data that produced a hockey stick covering 1000 years as opposed to data covering just the instrumental record.

  158. Chic Bowdrie says:

    Pekka,

    “The controversy is strongly linked to the existence of two different framings of climate change.”

    I totally second this analysis. I believe your comments about personality wars are also correct. However, I am not interested in personality wars but rather getting to the truth about how temperatures 1000 years ago compare to today. What motivated Mann (and others?) to use a particular methodology is secondary. The primary question is whether the methodology is appropriate.

  159. John Hartz says:

    Readers of this comment thread will want to check out:

    The Climate Deniers’ Newest Argument,

    an Op-ed by Jeffrey Kluger posted by Time magazine yesterday Sep 29, 2014.

    The lede of Kruger’s critique of Steven E. Koonin’s recent piece in the Wall Street Journal:

    It’s a lot easier to attack environmental scientists when you make up something they didn’t say—and then criticize them for saying it

  160. matt says:

    its 2014

  161. anoilman says:

    jsam: Indeed. That sure looks bad for Steve McIntyre. I wonder if he noticed the mistake.

  162. Chic,
    If you want to learn about past temperatures, you should look at the most up-to-date research, not the first attempts to find out. I’m not expert on that, but I would start at PAGES 2k.

  163. WebHubTelescope says:

    The case of the HackIntyre proves once again that man can not live on statistics alone. You need some physics for a balanced meal.

  164. Chic Bowdrie says:

    Pekka,

    The pages2k NAmerica2k working group published one paper (http://iopscience.iop.org/1748-9326/8/2/024008/article) exemplifying what the IPCC AR5 says about reconstructions:

    “Reconstructing NH, SH or global-mean temperature variations over the last 2000 years remains a challenge due to limitations of spatial sampling, uncertainties in individual proxy records and challenges associated with the statistical methods used to calibrate and integrate multi-proxy information (Hughes and Ammann, 2009; Jones et al., 2009; Frank et al., 2010a).” Furthermore, “The fundamental limitations for deriving past temperature variability at global/hemispheric scales are the relatively short instrumental period and the number, temporal and geographical distribution, reliability and climate signal of proxy records (Jones et al., 2009).” Specifically, IPCC AR5 notes “two further sources of uncertainty have been only partially considered in the published literature. First, some studies have used multiple statistical models (Mann et al., 2008)” … and secondly, “proxy-temperature relationships may change over time due to the effect of other climate and non-climate influences on a proxy, a prominent example being the divergence between some tree-ring width and density chronologies and instrumental temperature trends during the last decades of the 20th century (Briffa et al., 1998).”

    Allowing for the possibility that a set of tree ring data is sufficiently well correlated with the modern temperature record, and allowing for the possibility that the correlation is not coincidental, it still seems unclear whether Mann’s calibration procedure followed SOPs or, alternatively, was shown later to be a valid procedure. Apparently, this question has not been resolved as this post and recent posts at Climate Audit and Moyhu illustrate.

  165. John Mashey says:

    To avoid thread-diversion, thsoe interested in Dunning-Kruger discussions might visit Greg Laden’s post today

  166. Eli Rabett says:

    Chic, although statistical methods may be used to calibrate sensors, calibrating sensors is not a statistical idea, it is a physical one. In the case of proxies you not only have to calibrate each proxy, but you have to validate it. Not understanding that is indeed one of McIntyre;s problems.

  167. Mal Adapted says:

    Michael 2, to John Mashey:

    Your demonstration reveals that you know the words “Dunning-Kruger” but clearly do not understand the meaning thereof or you would not be carelessly calling it an “affliction”.

    My irony meter just broke. In their seminal paper Unskilled and Unaware of it: how difficulties in recognizing one’s own incompetence lead to inflated self-assessments, the authors explicitly analogize their eponymous effect to the neuropsychiatric malady of anosognosia:

    In the neurosciences, practitioners and researchers occasionally come across the curious malady of anosognosia…
    In this article, we proposed a psychological analogue to anosognosia. We argued that incompetence, like anosognosia, not only causes poor performance but also the inability to recognize that one’s performance is poor.

    A quick’n’dirty web search reveals that affliction is a common synonym for malady. Thus, there’s little doubt that Messrs K and D would be comfortable calling the D-K effect an affliction. There is substantial doubt, however, that Michael 2 has read the original paper.

  168. John Mashey says:

    Mal, do visit that thread at Greg Laden’s, there is more amusement.

  169. anoilman says:

    John Mashey: Makes me worry what bin you put my posts into. (Hopefully, ‘best with beer’.)

  170. Pingback: Things | …and Then There's Physics

  171. Nick says:

    McIntyre stuck in a rut? He needs some mountain air

  172. Mal Adapted says:

    John Mashey:

    Mal, do visit that thread at Greg Laden’s, there is more amusement.

    Heh. “Great minds” and so forth. Quoth Frank Zappa, another true, twisted genius of our time:

    It’s not getting any smarter out there. You have to come to terms with stupidity, and make it work for you.

    Who knew the fossil-fuel-billionaires-who-shall-remain-nameless were Zappa fans?

  173. Mal Adapted says:

    BTW, for those who haven’t seen it, a previous thread at Greg Laden’s is worth the time. Commenter Brainstorms posted the best rebuttal to an indefatigable pseudo-skeptic that I’ve read, and I was pleased to tell him so.

  174. anoilman says:

    There has been a lot of follow up work on Dunning-Kruger by the way. Cultural biases are abound. North Americans are most vulnerable. Here if we don’t like something, we skip it, move on, or try something else. We choose something more to our liking. In Japan if they don’t understand something, they persist in trying to understand it.

    Here’s Fool’s Dilemma which was a radio broadcast;

  175. John Mashey says:

    Mal:

    Scientist/educators* like Ben Franklin and Thomas Jefferson thought it was a good idea for people to be well-educated, as did the states who set up land-grant universities around US, and many others.

    But it hassling been the case, well before Zappa, that it is advantageous for some people to avoid having people get educated.
    If your model of the world is~feudal:
    A) royalty who runs things
    B) serfs
    C) nobility who support A) among other things because they don’t want to be in C)

    You don’t really want B) to get too educated , although you may pick off really talented ones and make them part of C). That seems the function of entities like the Institute for Humane Studies, Chaired by Charles Koch…

  176. David Young says:

    Given the climate ball intensity of this issue, this might be a waste but for those who are even a little open minded, McIntyre does have a substantive general critique of paleoclimatology and its use of proxies. It’s not just about Mann even though its convenient perhaps to say it is. Apparently, Pages2K is updating their data and methods recently in response to some of these issues.

    This is a 15 year old controversy and Pekka has perhaps nailed the issue. The original paper was not really of high enough quality to warrant the huge play it was given by the community. Early papers are sometimes flawed and that’s part of the give and take of science. McIntyre is continuing to do new work, but this thread seems to me to contain nothing new at all.

  177. DY,

    Early papers are sometimes flawed and that’s part of the give and take of science.

    If more people acknowledged/accepted that, the tone of this discussion (in general I mean) would improve.

    McIntyre is continuing to do new work, but this thread seems to me to contain nothing new at all.

    I don’t think there was any claim that there was going to be new work in this thread.

  178. jsam says:

    I, for one, look forward to McIntyre’s peer-reviewed global temperature reconstruction paper.

    Or, as the 3rd Viscount Monckton of Brenchley might put it, “porcos ut volaret”.

  179. Andrew Dodds says:

    jsam – As long as it doesn’t arrive in the same month as Mr Watt’s Surface Stations paper. The excitement would surely overwhelm the blogosphere.

  180. John Hartz says:

    David Young: The OP is to Climateball what highlight videos are to sports.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s