Knowlege Laundering

Charlie Martin is looking through some of James Hansen’s emails and found this:

[For] example, we extrapolate station measurements as much as 1200 km. This allows us to include results for the full Arctic. In 2005 this turned out to be important, as the Arctic had a large positive temperature anomaly. We thus found 2005 to be the warmest year in the record, while the British did not and initially NOAA also did not. …

So he is trumpeting this approach as an innovation?  Does he really think he has a better answer because he has extrapolated station measurement by 1200km (746 miles)?  This is roughly equivalent, in distance, to extrapolating the temperature in Fargo to Oklahoma City.  This just represents for me the kind of false precision, the over-estimation of knowledge about a process, that so characterizes climate research.  If we don’t have a thermometer near Oklahoma City then we don’t know the temperature in Oklahoma City and lets not fool ourselves that we do.

I had a call from a WaPo reporter today about modeling and modeling errors.  We talked about a lot of things, but my main point was that whether in finance or in climate, computer models typically perform what I call knowledge laundering.   These models, whether forecasting tools or global temperature models like Hansen’s, take poorly understood descriptors of a complex system in the front end and wash them through a computer model to create apparent certainty and precision.  In the financial world, people who fool themselves with their models are called bankrupt (or bailed out, I guess).  In the climate world, they are Oscar and Nobel Prize winners.

Update: To the 1200 km issue, this is somewhat related.

  • Mark

    I can see where some areas of the earth can be modeled from data rather far away. But you have to be really selective otherwise you have problems with the data.

    For instance the summer foggy weather in San Francisco is not at all an indicator of the weather in Reno, only 200 miles away, and the winter temps in Reno (snow and mid- 20 temps) have little relation to the Mid 60 temps in San Francisco during similar periods.

    San Diego is similar in summer, the temp in Downtown SD has little bearing on the temp in El Cajon just 15 miles away. It rarely gets above the 70’s in downtown SD, but in El Cajon, it can be 100 degrees, one day 80 the next while the downtown SD temp stays about the same. Maybe you could figure out a fudge factor, to figure when downtown is an average of 75 El Cajon is an average of 90 – but by the time you do all the research and temp studies to make that work – you may as well just use the temp readings from El Cajon. (I don’t know if the weather service actually officially measures there – just using it as an example)

    I would be surprised if there were a really good correlation between London and Edinburgh as someone suggested above.

  • papertiger

    @ Steve E

    It’s a nice notion to think you can avoid the pests by not reading comment sections at blogs, but look a little closer at the “scientific” paper I linked.

    It’s basicly a reiteration of Raper et al. (1984). They take two stations, one at the tip of the Antarctic peninsula (Faraday) and another on an island in the Atlantic (Orcadas), both outside the Antarctic Circle, then boister them with statistical bullshit of other stations that have supposed warming smaller then the margin of error.

    Which leaves the question, why did Phil Jones and crew even bother to write it?

    The only answer, to refute Doran et al. (2002) which used their own HadCRUT data set and found a net cooling of the entire continent.

    The rats are running the system. The Orkin man will not be called. We are on our own.

  • Wally

    Waldo

    “You would forgo the treatment, even though every Dr says you should take it? What is your reasoning behind this?”

    Well gee, as we delve further and further into hypothetical world, if every doctor wanted me to go through some treatment that appeared to be the equivalent of blood letting (something every doctor would have suggested a few hundred years ago), I wouldn’t do it. Medicine is better now, and grounded on good science, but that’s the point. I want to be able to see the science and be convinced of it before I undertake a possibly life threatening or altering treatment. Maybe for you doing such is too difficult, but it isn’t for me. In the end doctors are human, they make mistakes, and as the patient you and I are the first line of defense against mistakes. But to be able to help detect and prevent mistakes, you have to be an informed and thinking patient. If you want to close your eyes and pray everything your doctor tells you is 100% true and applicable to you specifically, fine by me. I got no skin in that game.

    “No. I addressed the argument:

    (march 15 12.23): ‘I don’t have a case on this stuff and I never claimed to. kuhnkat’s criticisms could be worthy (and hence should def. be seen) but I don’t know.”

    Only this was after I called you out for blindly claiming that the data does exist which Kuhnkat asked for yet failing to give a reference and stating that he should just publish his work, which implied and you latter even specified, that his argument here is not going to “change” anything. Thus the red herring. You attacked the location of the argument being presented, not the argument itself. Which all goes to proving the point that you aren’t here to discuss the science. You’re just here to argue.

  • Wally

    Hunter (the AGW one),

    “Perhaps you misunderstood, or perhaps you didn’t read the paper properly. The correlation coefficient at mid to high latitudes is almost 1 for nearby stations, and drops to 0.5 at an average distance of 1200km.”

    I did get that actually, sorry if I wasn’t specific enough. Regardless they say temp anaomolies at mid-high lat. are highly correlated up to 1000 km away. If they stuck to near by stations, maybe 200 km, and inside a correlation coef. of >.7-.8, I’d be with them. But they didn’t. They are claiming a correlation coef. of about .5-.6 is high. I don’t buy that. Plus they don’t give you a measure of their confidence in that correlation coef.

    “The weighting assigned to a station’s recorded temperature, when calculating a temperature for a given spot, varies linearly with distance from 1 at 0km, to 0 at 1200km.”

    Yes, and at 1200 km the weighting goes to zero. However, this doesn’t solve the problems I have.

    They say:
    “The smoothed global temperature increasebsy about .5øC
    between 1880 and 1940, decreases by about 0.2øC between
    1940 and 1965, and increases by about 0.3øC between 1965
    and 1980. The northern hemisphere temperature change is
    rather similar to the global change, increasing by 0.6øC
    between 1880 and 1940, decreasing by 0.3øC between 1940
    and 1970, and increasing by 0.3øC between 1970 and 1980.”

    Where is a standard devation or confidence interval and the P value for these .3 degree changes over 15 years being significantly different from zero? They have gone through a lot of statistical gymnastics here. Finding the correlation coef. (not giving a confidence) and then creating a weighting equation based on that result, but they don’t state the error right there with the paper’s main result? I can only guess the reason. Later in the paper they bring up the SD in just your average annual measurement of the temp from the norms (which gets as high as 1 degree C, when we’re talking .3ish over 15 years!), and errors in other models.

    Then lets look at figures 5 and 6. They actually show us some error, in this case 95% Confidence interval. Now look at the size and placement of them. They have 5 CI in the firgure, and of course they put them in the peak or trough years instead of just creating a year by year top and bottom CI through the whole graph. This is highly suspect. Why did they only pick those years to show me their error? And did they even do a test to see if which years are actually significantly different from eachother? Looking at the 1880-1900 warming, that looks like maybe its significant. The 1900-1940 looks fairly convincing, but again can you give me a test? Then the last warming from 1960-1980 again looks like it might be significant, but how about a test? I’m left feeling like there might be a heck of a lot of cherry picking going on here. What do the CI look like in years they didn’t care to report it? And they want to assign this .3 degree warming from 1970 to 1980, well fit that and tell me what the error is! Then of course, the whole while this is the error in the 5 year mean and the annual error in they call the GCM, not their actual model. So they don’t have a direct measure of their model’s error ANYWHERE!

  • Steve E

    papertiger
    “Which leaves the question, why did Phil Jones and crew even bother to write it?
    The only answer, to refute Doran et al. (2002)…”

    Sadly, I think you’re right. We see evidence of this type of tactic (create doubt, discredit) in the Climategate emails.

  • papertiger

    @ shills

    Where else would an assistant file clerk point to when in doubt?

    Here’s a bit on the character of Schmidt science.

    Part one: How scumbags do science.
    Part two: Why scumbags do science.

    Not surprised that you would run there shills.

  • Shills

    @ Hunter (denialist)

    Skeptics, unlike true believers, actually check stuff out.
Here is Hansen’s interview by Rob Reiss, an AGW promoter who totally backs your faith:

    Yeah, I saw that page too. What is your point?

    @ Wally:

    You say: ‘I want to be able to see the science and be convinced of it before I undertake a possibly life threatening or altering treatment’

    Kinda scary when you put the whole life threatening thing in there. The more accurate hypothetical is that doing nothing is gonna cost you more harm than doing the treatment (which is not life changing). The focus is on the deferment to authority, when you lack an understanding. If you wouldn’t do that, then why?

    You say: Only this was after I called you out for blindly claiming that the data does exist which Kuhnkat asked for yet failing to give a reference and stating that he should just publish his work

    Pays to know the context of a discussion. To address Kuhnkats issue I don’t need a paper ref. because, the data(all climate data) that all these papers come from covers much greater time periods than an ENSO cycle. And even if all this stuff was absent, and i just said ‘i dunno’ that would be addressing the argument. Besides, the fault of a red herring is not the claiming of unreferenced claims but the attempt to support a conclusion by an intentional diversion to irrelevant premises. No red herring.

    See:

    Kat said: ‘where is the empirical study that covers an entire ENSO cycle showing the relationship of multiple pairs of stations with distances of up to 1200 kilometers separating them??’

    I said: ‘The data comes from time periods greater than ENSO cycles. If the methods used still seem doubtful (I haven’t a clue) I suggest you do something about it. Submit a paper.’

    I didn’t even claim that I had a paper, which you think i did.

    @ Papertiger:

    YOu say: ‘Not surprised that you would run there shills.’

    Lol. dude if you don’t like realclimate, well ok. But it was just meant as a link to the paper being looked at.

  • hunter

    Shills,
    Your question is an excellent demonstration of what true denialism is.
    When I asked if you were OK with Hansen lying outside of peer review, you claimed to doubt the source, but did not answer the question.
    I provided the quote, and the source of the quote, and you now ask what is the point.
    You are immune to new information,a nd have chosen to suppress your critical thinking skills, if you in fact have any.
    You know there is something there, because you claimed to doubt the source.
    Now that you know it is real, you simply claim it doesn’t matter.
    And it doesn’t matter- to someone in deep denial.

  • Shills

    @ Hunter (denialist):

    you say: ‘When I asked if you were OK with Hansen lying outside of peer review, you claimed to doubt the source, but did not answer the question.’

    I did answer your question i said ‘no’.

    you say: ‘I provided the quote, and the source of the quote, and you now ask what is the point.’

    Lol. You seem to miss the point. I knew already that the hearsay was from an interview:

    I said way before: (march 15, 10.52) ‘Hansen’s manhatten line, was from an interview, not science lit.’

    Then you gave me the link i’d already seen, which doesn’t address my science-lit. concerns. I still maintain that hearsay of a conversation in an interview isn’t worth much.

  • hunter

    “If they stuck to near by stations, maybe 200 km, and inside a correlation coef. of >.7-.8, I’d be with them. But they didn’t. They are claiming a correlation coef. of about .5-.6 is high. I don’t buy that. Plus they don’t give you a measure of their confidence in that correlation coef.”

    This is just tiresome noise. You complain about things without giving any sensible or rational reasons why. If you want to be taken more seriously, then how about you quantify exactly what effect a different approach would have. What would global temperature trends look like if, as you suggest, the weighting given to stations dropped to zero at 200km instead of 1200km? I’d like to know and so, I’m sure, would the scientific community.

    With all your whinging about errors not being quantified it seems you have overlooked section 5 of the paper, which runs from page 16 to page 25, and is entitled “error estimates”.

  • Wally

    Shills,

    >‘where is the empirical study that covers an entire ENSO cycle showing the relationship of multiple pairs of stations with distances of up to 1200 kilometers separating them??’

    I said: ‘The data comes from time periods greater than ENSO cycles. If the methods used still seem doubtful (I haven’t a clue) I suggest you do something about it. Submit a paper.’

    I didn’t even claim that I had a paper, which you think i did. <

    No I didn't think you had a paper. This is the point. Kat asked for one and you just said (paraphrasing), "yes there is one." And you wonder why I think you're full of shit?

  • Wally

    Hunter,

    “You complain about things without giving any sensible or rational reasons why. If you want to be taken more seriously, then how about you quantify exactly what effect a different approach would have. What would global temperature trends look like if, as you suggest, the weighting given to stations dropped to zero at 200km instead of 1200km? I’d like to know and so, I’m sure, would the scientific community.”

    Uh, my point was that .5 is not a “high” correlation coef. as claimed. If .5 is classified “high” what is .6, or .9, or .99?

    >With all your whinging about errors not being quantified it seems you have overlooked section 5 of the paper, which runs from page 16 to page 25, and is entitled “error estimates”.<

    Sounds like someone wasn't actually reading the paper and my post. I directly mentioned things brought up only in section 5. Such as the 95% CI presented in figure 6 was from the GCM and not from their actual model (which if you missed it is that nice cute equation where the weighting goes to zero at 1200km). If you read section 5 what you will find is that they don't have give the error of their model in anyway what so ever. They try to get around that by giving you the error in the GCM or just yearly measurements relative to the historic mean. They also don't assign any level of confidence that their 3 degree warming trend from 1965 to 1980 was even significantly different from zero.

    Now if you think they did, give me the numbers and maybe some text from the paper. What was the SD or CI in those tends they saw that I quoted above and where they significantly different from zero?

    No paper in my field could be published without answering those simiple questions….so it must be in there right….somewhere….

  • Wally

    Hunter (the sane one)

    “When I asked if you were OK with Hansen lying outside of peer review, you claimed to doubt the source, but did not answer the question.
    I provided the quote, and the source of the quote, and you now ask what is the point.”

    This seems to be Shills’ MO. First, at all costs, avoid the actual argument or data. Second, if possible, attack the author or the source. And once completely boxed in, argue about the point.

  • Shills

    @ Wally:

    you say: ‘Kat asked for one and you just said (paraphrasing), “yes there is one”‘

    wow, that’s some pretty liberal paraphrasing. how’d you get that from:

    ‘The data comes from time periods greater than ENSO cycles. If the methods used still seem doubtful (I haven’t a clue) I suggest you do something about it. Submit a paper’.

    You think my ‘the data’ line is referring to a paper? No. I simply make the claim that the data would cover periods of time much longer than an ENSO cycle (By necessity, all climate data would cover something like almost three decades, I think).

  • hunter

    Shills, You are true to your name.

  • Shills

    Shills,

    “You think my ‘the data’ line is referring to a paper? No. I simply make the claim that the data would cover periods of time much longer than an ENSO cycle (By necessity, all climate data would cover something like almost three decades, I think).”

    Sure the data exists or could exist, but the question that was asked was, “where is the empirical study that covers an entire ENSO cycle showing the relationship of multiple pairs of stations with distances of up to 1200 kilometers separating them??”

    So, whether or not you were talking about just the data or an actual study (which is what he asked for) doesn’t really matter, because you didn’t provide that information. Also, if you are simply refearing to data existing, what does that data actually show? These were the kind of things kuhnkat was asking for, and your responce failed completely to address those things.

  • Wally

    Ah man, that last post was me. Apperently I had a little confussion in what was my name and who I was addressing…

  • Shills

    @ Wally:

    As I said earlier. Saying ‘I dunno’ addresses the question. No red herring.

  • Wally

    Shills,

    No, you didn’t do that until you were pressed to stay on topic.

    Kuhnkat said: “where is the empirical study that covers an entire ENSO cycle showing the relationship of multiple pairs of stations with distances of up to 1200 kilometers separating them? It doesn’t exist….”

    You said: “The data comes from time periods greater than ENSO cycles. If the methods used still seem doubtful (I haven’t a clue) I suggest you do something about it. Submit a paper.”

    You didn’t say this until I pointed out how you attempted to lead us down this red herring: “kuhnkat’s criticisms could be worthy (and hence should def. be seen) but I don’t know.”

    Now, if you want to talk about the science, and get past this “I dunno, why don’t you do something about it” BS, that would be great. The other hunter was willing to do that, but has since disappeared. I’m not convinced he really knew was a confidence interval was, maybe that’s why he disappeared. How about you?

  • Shills

    @ Wally:

    I said something to that effect originally: ‘(I haven’t a clue)’. No red herring.

    You say: ‘Now, if you want to talk about the science, and get past this “I dunno, why don’t you do something about it” BS, that would be great.’

    You are the one bringing up these insignificant logical fallacy accusations. If you don’t want to defend your accusations than keep them to your self next time.

    Lets talk about the science. How many times have I asked you to show peer-reviewed evidence against AGW theory and you just ignore the request or make some other comment (red herring maybe?) about burden of proof (which I address),or over-blow the IPCC’s errors.

  • hunter

    “Uh, my point was that .5 is not a “high” correlation coef. as claimed. If .5 is classified “high” what is .6, or .9, or .99?”

    That is not claimed. You don’t seem to be able to understand that the weighting drops to zero at 1200km, which is the distance at which the correlation drops to ~0.5. You don’t yet understand the point of the paper or the basic methodology.

  • Wally

    Shills,

    “I said something to that effect originally: ‘(I haven’t a clue)’. No red herring. ”

    This was about the methods of some unknown paper or study, not about the existence of the paper or study, which was his question. So yeah, still a red herring.

    “You are the one bringing up these insignificant logical fallacy accusations. If you don’t want to defend your accusations than keep them to your self next time.”

    Rather comical response from you, haven’t I been defending these accusations the whole time and continue to do so? Also, while you may think this is “insignificant” (as you apparently think everything said in these comments is anyway), I disagree. You can’t answer a specific question about the existence of the study in question without diverting onto tangential points. So who’s the one willing to talk science again?

    “How many times have I asked you to show peer-reviewed evidence against AGW theory and you just ignore the request or make some other comment (red herring maybe?) about burden of proof (which I address),or over-blow the IPCC’s errors.”

    So, you just want any anti-AGW paper out of the blue? You don’t want to actually talk about things already being discussed? Uh, and you wonder why you get accused of bring up red herrings? Hunter and I are already talking about the paper in question, read it, join us.

  • Wally

    Hunter,

    First, I notice you again failed answer my question. If you need a reminder: “What was the SD or CI in those tends they saw that I quoted above and where they significantly different from zero?”

    Ultimately, this is the most important problem with the paper. They make this model, find various warming/cooling trends over different times. But they don’t actually report the error of their model, nor do any kind of test to determine if their model’s results are significantly different from zero. If you think they do, please, show me where.

    Now about the “high” correlation, this is in the abstract: “The temperature changes at mid- and high latitude stations separated by less than 1000 km are shown to be highly correlated”

    At 1000km, and latitudes above 64.2N the correlation coef is .6.
    At 1000km, and latitudes between 44.4 and 64.2N is somewhere between .6 and .5. Hard to tell from fig 3.

    They even state: “For example, in these regions the average correlation coefficient for 1000-km separation was found to be within the range 0.5-0.6 for each of the directions defined by 45 ø intervals.”

    They go on: “The 1200-km limit is the distance at which the average correlation coefficient of temperature variations falls to 0.5 at middle and high latitude”

    So they are certainly talking about correlation coefs of .5-.6 when the say “highly correlated” in the abstract. It is true that the closer distances in this <1000km are highly correlated, but anything more then about 500-700 km, especially at those "mid" latitudes are certainly NOT highly correlated.

  • hunter

    “So they are certainly talking about correlation coefs of .5-.6 when the say “highly correlated” in the abstract. It is true that the closer distances in this <1000km are highly correlated, but anything more then about 500-700 km, especially at those "mid" latitudes are certainly NOT highly correlated."

    And that's why the weighting drops to zero. What's hard to understand here? It's all pretty ridiculously straightforward but you seem to be struggling. The words in the abstract seem to be making you misunderstand the actual methodology, and it doesn't seem that you're willing to learn how to correct that misunderstanding. Arguing over the meaning of the word 'highly' is irrelevant and tiresome.

  • Wally

    Hunter,

    “And that’s why the weighting drops to zero.”

    But the weighting is still there despite a pretty middling correlation coef.

    “Arguing over the meaning of the word ‘highly’ is irrelevant and tiresome.”

    Hmm, but that’s all you seem willing to talk about. Why haven’t you addressed my other critiques?

  • Shills

    @ Wally:

    You say: ‘this was about the methods of some unknown paper or study, not about the existence of the paper or study, which was his question. So yeah, still a red herring.

    Lol. It was about both. It was about the effects of ENSO cycles on the methods being used in some study which may or may not have being addressed in another paper.

    He asked for answers, spec, a paper, that addressed the issue. His issue is with the ENSOs affecting the method’s validity. I suggest an answer (the long time intervals, which would expose such patterns). And, with no other ideas, and no papers that address his issue (hence I have no ‘clue’),–and the fact that he has already declared there to be no papers– I suggest he submits one himself (what else would one do if they had found a prev. unknown error). His issue is addressed. No red herring.

    You say: ‘haven’t I been defending these accusations the whole time and continue to do so?’

    Lol. Then why were you insisting that we leave the ‘BS’ topic? You Wally, are a comic genius.

    You say: ‘So, you just want any anti-AGW paper out of the blue?’

    It doesn’t seem to matter when I want it, you don’t deliver. besides, the times I ask you for evidence in this thread are merely a continuation of the same request you just ignore from the ‘Just Your Typical Interview on Scientific Issues..” thread. And that request was very much in line with the discussion. Yes, I’m still waiting…

  • Wally

    Shills,

    “It was about the effects of ENSO cycles on the methods being used in some study”

    Ah, right. It was about the methods in some unidentified paper…that kuhnkat wanted to see…that you couldn’t provide, but you knew the methods….

    “Then why were you insisting that we leave the ‘BS’ topic?”

    Not because I don’t want to or won’t defend my argument, but because we’re going in circles. Its been explained to you several times how you committed a red herring. If you don’t believe that, fine. I suppose you could choose not believe 2+2=4 as well. I can’t make you understand something, I can only explain it.

    “It doesn’t seem to matter when I want it, you don’t deliver.”

    This is a lie. We’ve talked about studies that dispute AGW before. Did I not give you a list 500 papers for you to skim through and read at your pleasure? Did we not discuss more then one of them. And of course you picked the most trivial ones, and attacked the location of publications instead of the actual papers….

    “the times I ask you for evidence in this thread are merely a continuation of the same request you just ignore from the ‘Just Your Typical Interview on Scientific Issues..” thread. And that request was very much in line with the discussion. Yes, I’m still waiting…”

    I had left that thread, as I can’t continually, day after day, pour several hours into your nonsense, but looking back ADiff, pointed you in the direction of our desired papers, but you seem unwilling to lift a single finger to find them. If you can’t do that, how much does that say about your ability to honestly read and critique a paper?

    Now if you really want a paper to discuss, I can give you one (or many), but before I do so, I would like you to answer this question: Why are you not willing to engage in a discussion of the paper that is the topic of this discussion?

    Now, how about we read this: http://www.agu.org/journals/gl/gl0916/2009GL039628/
    No attacking authors or place of publication, just deal with what is said in the paper.

  • Shills

    @ Wally:

    You say: ‘Ah, right. It was about the methods in some unidentified paper…that kuhnkat wanted to see…that you couldn’t provide, but you knew the methods….’

    No. It was about the neg. effects the ENSO cycle might have on the methods being critiqued at the start of this thread. No red herring.

    You say: ‘but because we’re going in circles. Its been explained to you several times how you committed a red herring. If you don’t believe that, fine. I suppose you could choose not believe 2+2=4 as well. ‘

    Lol. Dito, somewhat.

    You say: ‘This is a lie. We’ve talked about studies that dispute AGW before. Did I not give you a list 500 papers for you to skim through and read at your pleasure?’

    No, you are mistaken. But I know the list you are talking about. If that is your best evidence than fine, I’ll take it.

    You say: ‘ but looking back ADiff, pointed you in the direction of our desired papers, but you seem unwilling to lift a single finger to find them’

    Lol. I searched through a whole heap of the posts on the front page. No ref. peer-reviewed papers. And ADiff has only replied back in the last few days. My question went unanswered for ages.

    You say: ‘Why are you not willing to engage in a discussion of the paper that is the topic of this discussion?’

    Never said I wasn’t willing. I figured Hunter had a better grasp of this stuff and he was addressing your questions.

    I can’t read that paper you linked to; Non subscriber.

  • Wally

    Shills,

    “No. It was about the neg. effects the ENSO cycle might have on the methods being critiqued at the start of this thread. No red herring.”

    But how do you know what “the neg. effects the ENSO cycle might have on the methods being critiqued” if you can’t produce the study that contains the methods?

    “And ADiff has only replied back in the last few days. My question went unanswered for ages. ”

    But ADiff posted and you responded to him without looking for the papers he at least pointed you towards. Your question being unanswered for “ages” is irrelevent, hyperbolic, and (GASP) a red herring.

    “I figured Hunter had a better grasp of this stuff and he was addressing your questions.”

    So, far he’s failed to address the single most important question anyone should have after reading this paper, are their findings significantly different from zero, and if so, by how much? Now if you, or anyone, could point me toward that information, that would be great.

    “I can’t read that paper you linked to; Non subscriber.”

    Well, this is likely to be a problem with many papers I might pick, so why don’t you pick something. Also, why don’t you just pick something by Lindzen or Pielke, so to make sure you don’t attack the authoer’s credibility. Though if you don’t have access to journals requiring subscribership, we probably can’t get around picking papers published in journals you would attack. Which would of course limit the numbers of papers with authors you find credable as well….

  • Shills

    @ Wally:

    You say: ‘But how do you know what “the neg. effects the ENSO cycle might have on the methods being critiqued” if you can’t produce the study that contains the methods?’

    I don’t know about any negative effects, it was Kat’s assertion that they existed. And he wanted a paper that might cover these neg. effects on the methods being described by Meyer at the very beginning.

    You say: ‘But ADiff posted and you responded to him without looking for the papers he at least pointed you towards. Your question being unanswered for “ages” is irrelevent, hyperbolic, and (GASP) a red herring.’

    ehh, no. I said I looked through the front page and found nothing. ADiff says ‘presentations’ and ‘publications’. The only one I know of is his movie. Where do I find these publications? Re. Question going unanswered: I said this to indicate that you didn’t just not reply because someone else was doing the job, which could have been inferred from reading your post (but not saying that you implied it).

    You say: ‘Though if you don’t have access to journals requiring subscribership’

    I have dodgy Uni library subscription. It doesn’t nec. work for everything, but I’ll try your paper again. And I’ll try Hunter’s.

  • hunter

    “But the weighting is still there despite a pretty middling correlation coef.”

    “Zero”, as most small children would be able to tell you, does not equal “still there”. How about I say it again: the weighting drops to zero at 1200km, the distance at which the correlation between station anomalies drops to 0.5 on average. “pretty middling” is a meaningless statement. You seem unable to state clearly exactly what you are objecting to and why.

    “Why haven’t you addressed my other critiques?”

    Did you make any? I’ve only seen idiotic quibbling over the meanings of words, such as what I just quoted. If you had any scientific questions I missed them. I see no sign that you even understand the methodology so you’re a long way from being able to critique it.

  • Wally

    Hunter,

    >“Zero”, as most small children would be able to tell you, does not equal “still there”. How about I say it again: the weighting drops to zero at 1200km, the distance at which the correlation between station anomalies drops to 0.5 on average.What was the SD or CI in those tends they saw that I quoted above and where they significantly different from zero?

    Ultimately, this is the most important problem with the paper. They make this model, find various warming/cooling trends over different times. But they don’t actually report the error of their model, nor do any kind of test to determine if their model’s results are significantly different from zero. If you think they do, please, show me where.<

    Stop ignoring this question. Ignore the above issue of the corellation coef. if you like and just address this.

  • Wally

    Ok, part of my post got screwed up let me just fix it and repost:

    Hunter,

    >“Zero”, as most small children would be able to tell you, does not equal “still there”. How about I say it again: the weighting drops to zero at 1200km, the distance at which the correlation between station anomalies drops to 0.5 on average.<

    Wow, ok, I guess you failed to connect the lines here. Something dropping towards zero, is different from just zero. My problem is realying on anything with such a middling coef. even if it is of low weight. All that does is introduce more error into your system as you increasingly rely on a few measurements. Which ties into the larger point. They create this model, but they don't even both to mention the error of their model. They only mention the varience in yearly readings from the historic mean or the error in the GCM, which isn't their model.

    Which all leads into this:

    "Did you make any? I’ve only seen idiotic quibbling over the meanings of words, such as what I just quoted. If you had any scientific questions I missed them."

    You missed them? That's a laugh, it was the first thing I wrote about two posts ago to you, and what I spent most of my time on in my more thorough critique of the paper. Let me restate it:

    "What was the SD or CI in those tends they saw that I quoted above and where they significantly different from zero?

    Ultimately, this is the most important problem with the paper. They make this model, find various warming/cooling trends over different times. But they don’t actually report the error of their model, nor do any kind of test to determine if their model’s results are significantly different from zero. If you think they do, please, show me where."

    Stop ignoring this question. Ignore the above issue of the corellation coef. if you like and just address this. Without answering this question, this paper is worthless.

  • hunter

    “My problem is realying on anything with such a middling coef. even if it is of low weight”

    For fuck’s sake. “Middling” is a stupid, subjective, meaningless and unscientific word, that means nothing at all in this context. Your inability to grasp this is pathetic. If you were interested in the science you would get hold of the data, apply a model that you like better, and see what happened. But you’re a tiresome fuckwit so you just repeat the same shit again and again without making any effort to investigate. Explain quantitatively what weighting scheme you would prefer, and why it would be superior, or shut up.

    The rest of the paper is largely irrelevant to the discussion here, which is about the observational fact of correlations between temperature anomalies over large distances. I suggest you read again the lengthy section entitled “Error estimates”, and if you don’t find the answer you want then contact the lead author. His e-mail address is readily available. I’m sure he’ll appreciate your “critiques”.

    Regarding the 1200km issue, you have yet to do anything more than idiotically quibble about the meaning of words. If you ever do think of any scientific questions about it, then get back to us.

  • Wallly

    Interesting that you still can not tell me: “What was the SD or CI in those tends they saw that I quoted above and where they significantly different from zero?”

    Its also interesting that you’ve had to rely on some rather uncreative four letter word spin offs to make an “argument.”

    I’m done with you.

  • Wally

    Ok, I lied. One last thing:

    “For fuck’s sake. “Middling” is a stupid, subjective, meaningless and unscientific word, that means nothing at all in this context.”

    But the other’s used a similar subjective quantifier in their abstract, that being “high”, haven’t see you call them a “fuckwit.” It is also similarly “unscientific” to report a finding from an algorithm, but not give the algorithm’s error.

    Anyway, nice talking to you.

  • hunter

    Yeah, they went on to quantify the correlation in quite some detail. All you seem to be able to do is describe the correlation near 1200km as “middling”. You lack the intelligence to quantify your qualitative statements, which thereby do not mean anything. If you want to say something meaningful, explain what weighting scheme you’d prefer, and why it would be better.

    As for your other questions, you should read the section of Hansen and Lebedeff 1987 entitled “Error estimates”, and also the many subsequent GISS papers further describing their methodology and results. If you read them with an open mind and a modest amount of scientific ability you’ll find what you seek. I don’t think you have either of those things though.

  • Wally

    “As for your other questions, you should read the section of Hansen and Lebedeff 1987 entitled “Error estimates”, and also the many subsequent GISS papers further describing their methodology and results.”

    Read it. Point me to where they actually report the error in their model.

  • hunter

    You don’t have an alternative idea for a weighting scheme, or any coherent explanation for why you don’t like the Hansen and Lebedeff one, you’re only capable of saying you don’t like it. That’s a very weird prejudice to have. I wonder what other weird prejudices you have.

    The answers you seek regarding errors are in the section called “Error estimates”.

  • Wally

    Hunter,

    “You don’t have an alternative idea for a weighting scheme”

    I don’t need an alternative to determine if the one they create is worth a crap. Someone could create a model for Zeus throwing lightning blots to earth. I don’t need to create an alternative model for Zeus’ lightning blots to determine that their model is BS.

    “or any coherent explanation for why you don’t like the Hansen and Lebedeff one, you’re only capable of saying you don’t like it.”

    Not true. They find the average correlation coef. between stations at different distances and use that to make a weighting factor to extrapolate temperature measurements over unmeasured territory. However, they do not even begin to explain their confidence in that average correlation coef.

    Think of it like this: If I had two data points, (1,1) and (2,2). I could draw line between the two and say it has an R^2 of 1. But I only have to data points, so despite that high R^2 I have very low confidence in that R^2 and the resulting line between the points. As I gather more data say (1.5, 1.5), (1.2, 1.2), (1.9, 1.9) my confidence in that trend line and the regression goes up. They don’t report that measurement of confidence.

    But it gets worse. They aren’t just creating one line here. They are presumably making thousands and thousands of measurements at a variaty of stations over time, then comparing them between station A and B, A and C, B and C, etc. They then take the R^2 from all of these relationships and create a new average correlation based on distance. But they don’t even give you the error in that average correlation on top of not reporting the error in the individual correlations.

    What they have done is created a regression on top of another regression. This kind of thing is going to introduce a lot of error, but they don’t even talk about it! And please stop saying its just in the “error estimates” section. I read that section, I didn’t see it. If you think its in there, find it, post it.

    Basically, what they need in that figure with a line through all those correlation coef. dots vs. distance is error bars. Now those dots range from ~.4 to ~.95 at 500 km in the mid latitudes. So I’m guessing a 95% CI is going to be fairly sizable. Then the higher latitudes have a similar distribution, but with far less measurements, meaning even more error.

    “I wonder what other weird prejudices you have.”

    I wonder why you can’t actually find the error terms or confidence intervals in their actual analysis that I’m asking for? Is it because you don’t understand? Is it because you’ve never had even an intro stat class? Or is it your own biases, you know they aren’t there, and you just want to brush my criticisms off with “its in the error estimates section?”

    Here, I’ll give you one more chance. Read the paper. Find the error terms I want, or this conversation is over.

  • hunter

    Stop being so pathetic. Stop demanding to be spoonfed things that you can’t be bothered to look for or investigate yourself. Uncertainties are discussed extensively in the paper, and if you don’t find that discussion satisfactory, take it up with the paper’s author. If you knew the first thing about statistics, and how to calculate standard errors, your own statement of “thousands and thousands of measurements” would tell you something important. But you obviously don’t know these things.

    You have failed to explain what problem you have with the observed correlation between station anomalies. You have failed to explain what problem you have with the weighting scheme used to derive surface temperatures. You have failed to explain what impact you think it might have on derived surface temperatures. You have failed to understand basic statistics. You have failed to read the paper properly. Not a very impressive list, is it? Then again, you “had a little confussion in what was my name” earlier, so I think we can see what kind of intellect you are turning to this problem.

  • Stevo

    “Find the error terms I want, or this conversation is over”

    Talk about spitting the dummy! You sound like a small child whose mother won’t buy it the toy it wants.

  • Wally

    Hunter,

    I don’t see anything in your post that either directly adresses my critisisms (instead you just blindly claim I “failed” to do X, Y or Z) or gives me the error terms I did not see in the paper.

    We’re done here.

  • Wally

    Stevo,

    So, I suppose I’m supposed to find something for myself that isn’t there?

    How about I just make something up? I don’t think Hunter has actually read, much less understood the “error estimates” section, so he probably won’t notice anyway. Would that make you feel better?

    Ok how’s this. They say:

    “The smoothed global temperature increasebsy about .5øC
    between 1880 and 1940, decreases by about 0.2øC between
    1940 and 1965, and increases by about 0.3øC between 1965
    and 1980. ”

    And later in the paper they report the error of their model to be +/- 2 degrees C annually at the 95% confidence interval.

    Ok, there. Prove me wrong.

  • Stevo

    Er, what? You can’t be bothered to read the paper so you will make something up? I don’t think you’re doing your credibility a whole lot of good here. Looks like you didn’t notice the following passages:

    “The standard deviation σ for the global mean temperature decreases from about 0.07°C in the 1880s to about 0.02°C in the 1960s”

    “the principal features in the global and hemispheric temperature changes are real, in the sense that they are not artifacts due to poor spatial coverage of stations.The long-term global trends illustrated in Figure 6, i.e., the 1880-1940 warming, 1940-1965 cooling, and 1965-1985 warming, are much larger than the estimated errors”

  • Wallly

    Stevo,

    “You can’t be bothered to read the paper so you will make something up?”

    Are you hunter reincarnate? I read the paper, the things I was looking for weren’t there.

    “The standard deviation σ for the global mean temperature decreases from about 0.07°C in the 1880s to about 0.02°C in the 1960s”

    Ah, yes. But notice that’s the standard deviation for the GLOBAL MEAN TEMPERATURE. That is not a measurement of the error in their analysis.

    “The long-term global trends illustrated in Figure 6, i.e., the 1880-1940 warming, 1940-1965 cooling, and 1965-1985 warming, are much larger than the estimated errors”

    They are? What is the error in their measurement of the trends? Surely they must report that if they are to say such a thing.

  • Stevo

    Now you’re really struggling. The standard deviation in the estimated global mean temperature is obviously a measurement of the uncertainty in their analysis. If I measure the height of a mountain, and tell you what it is with an uncertainty estimate as well, no doubt you would wail “But that’s an uncertainty for the HEIGHT OF THE MOUNTAIN. That is not a measurement of the uncertainty in your analysis”. You would be making no sense, just as you are not now.

    Let’s imagine I measure this mountain again much later. I find that it’s bigger now, by an amount much larger than the uncertainties on my two estimates of its height. Clearly, you would not accept that the mountain had actually grown.

    Your mind is obviously closed.

  • Wally

    Stevo,

    “The standard deviation in the estimated global mean temperature is obviously a measurement of the uncertainty in their analysis.”

    No it is not. Their analysis went far beyond just taking measurements of temps. They took those measurements they hand and extrapolated them over unmeasured territory. Then they went even further then that. They created another statistic on the correlation coef over distance. That introduces more error then you would have just in the standard deviation of yearly global mean temps. These to things so completely different that the st. dev. of the global mean is completely worthless. You can’t just measure the error of one thing and use it as the error for something sorta-kinda related. Its complete crap. In any other field, you’d be laughed at if you tried to do this.

    “If I measure the height of a mountain, and tell you what it is with an uncertainty estimate as well, no doubt you would wail “But that’s an uncertainty for the HEIGHT OF THE MOUNTAIN. That is not a measurement of the uncertainty in your analysis”. You would be making no sense, just as you are not now.”

    Lets run with this analogy, shall we? The problem is that they aren’t trying to get at the height of just one mountain, nor is that the measurement of error they present. What would be more analogous is if they measured the height of mountains and came up with a correlation coef. of height vs. distance. Then they used that correlation rate to create a function to estimate the height of mountains in a given area and then used the global average st. dev. of the height of mountains in general as an estimate of the error in their estimate of the average high of mountains in a given area….wow. K that was fun. Now the question is why not just keep track of the actual error in your analysis? Its pretty simple to figure out the error in a trendline and 95% CI. Why didn’t they do that?

    “Clearly, you would not accept that the mountain had actually grown.”

    You do know mountains grow and shrink all the time right, and we can measure that. Anyway, your analogy is crap, and I suspect your knowledge of statistical analysis doesn’t even included a T-test.

  • Wally

    Stevo,

    Here’s just a little something from wiki: “It is thought that the plate tectonics of the area are adding to the height and moving the summit (of Mt. Everst) northeastwards. Two accounts suggest the rates of change are 4 mm (0.16 in) per year (upwards) and 3 to 6 mm (0.12 to 0.24 in) per year (northeastwards),[20][23] but another account mentions more lateral movement (27 mm/1.1 in),[24] and even shrinkage has been suggested.[25]”

    We can measure these things, but only to certain degree of confidence. Thus, two groups tried to measure it and came up with very different results. Its entirely possible both measurements above are right, they are just at two opposite ends of the error (or not, the may be much larger then ~30mm/year). Further evidence you don’t actually think things through and your knowledge is a bit lacking.

  • Stevo

    Oh god, you really don’t understand what their analysis was about, do you? They sought to estimate global mean temperature anomalies. They devised a method to do so. They quote values, and uncertainties. You are embarrassing yourself to claim that uncertainty in the analysis and uncertainty in the value resulting from the analysis are somehow intrinsically different. It’s hard to know where to start with someone so ridiculously wrong.

    “Its pretty simple to figure out the error in a trendline and 95% CI. Why didn’t they do that?”

    They did. You’ve been told what they said. Apparently, though, you’ve poked your eyes out and sewn your ears shut in a fit of anti-scientific reactionary pique.