Bishop Hill Keenan on McKitrick
Sep 29, 2014 Climate: Statistics Climate: Surface Doug Keenan has posted a strong critique of Ross McKitrick's recent paper on the duration of the pause at his own website. I am reproducing it here.
McKitrick [2014] performs calculations on series of global surface temperatures, and claims to thereby determine the duration of the current apparent stall in global warming. Herein, the basis for those calculations is considered.
Much of McKitrick [2014] deals with a concept known as a “time series”. A time series is any series of measurements taken at regular time intervals. Examples include the following: the maximum temperature in London each day; prices on the New York Stock Exchange at the close of each business day; the total wheat harvest in Canada each year. Another example is the average global temperature each year.
The techniques required to analyze time series are generally different from those required to analyze other types of data. The techniques are usually taught only in specialized statistics courses.
Assumptions
The calculations of McKitrick [2014] rely on certain assumptions. In principle, that is fine: some assumptions must always be made, when attempting to analyze data. A vital question is almost always this: what assumptions should be relied upon? The question is vital because the conclusions of the data analysis commonly depend upon the assumptions. That is, the conclusions of the analysis can vary greatly, depending upon the assumptions.
The problem with McKitrick [2014] is that it relies on assumptions that are wholly unjustified—and, worse, not even explicitly stated. Hence, I e-mailed McKitrick, saying the following.
The analysis in your paper is based on certain assumptions (as all statistical analyses must be). One problem is that your paper does not attempt to justify its assumptions. Indeed, my suspicion is that the assumptions are unjustifiable. In any case, without some justification for the assumptions, there is no reason to accept your paper's conclusion.
The issue here is not specific to statistics. Rather, it pertains to research generally: in any analysis, whatever assumptions are made need to be justified.
McKitrick replied, claiming that “The only assumption necessary is that the series [of temperatures] is trend stationary”. The term “trend stationary” here is technical, and is discussed further below.
There are two problems with McKitrick's claim. The first problem is that trend stationarity is not the only assumption made by his paper. The second problem is that the assumption of trend stationarity is unjustified and seemingly unjustifiable. The next sections consider those problems in more detail.
The assumption of linearity
McKitrick claimed that his paper only made one assumption, about trend stationarity. In fact, the paper also assumes that all the relevant equations (for the noise) should be linear. Hence, I e-mailed McKitrick back, saying the following.
Stationarity is not the only assumption. Your paper also includes some assumptions about linearity … I do not see how [linearity] can be justified….
McKitrick did not respond. Five days later, I sent another e-mail, again raising the problem of assuming linearity. This time McKitrick replied at length. His reply, however, did not mention linearity.
The climate system is nonlinear. This is accepted by virtually everyone who has done research in climatology. For example, the IPCC has previously noted that “we are dealing with a coupled non-linear chaotic system” [AR3, Volume I: §14.2.2.2]. Hence the assumption of linearity is very dubious. There might be occasions where it is suspected that a linear approximation is appropriate, but if so, then some argument for the appropriateness must be given.
The assumption of trend stationarity
For technical details of what it means for a time series to be trend stationary, see the Wikipedia article. This section considers issues that do not require those details.
McKitrick's first e-mail to me acknowledged that trend stationarity “makes an enormous difference for defining and interpreting trend terms”. Simply put, if the trend in global temperatures is not assumed to be trend stationary, then the calculations of McKitrick [2014] are not valid.
The abstract of McKitrick [2014] states that the calculations used in the paper are “valid as long as the underlying series is trend stationary, which is the case for the data used herein” (emphasis added). The emphasized claim seems to imply that trend stationarity of the temperature data is an established fact.
The body of the paper says that the temperature data is “assumed to be trend-stationary”. The paper makes no attempt to justify the assumption. At least, though, the body of the paper acknowledges that trend stationarity is an assumption, rather than a fact.
McKitrick's first e-mail to me said that “decisive tests [for trend stationarity] are difficult to construct”. Thus, McKitrick seems to be acknowledging that he has no decisive statistical tests to justify the assumption of trend stationarity.
McKitrick's first e-mail also referred to a workshop, held in 2013, at which “there were extended discussions on whether global temperature series are stationary or not”. Thus, this effectively acknowledges that McKitrick knows trend stationarity is nowhere near being an established fact.
McKitrick's second e-mail attempted some justification for assuming trend stationarity. It said this: “The reason I do not accept the nonstationarity model for temperature is that it implies an infinitely large variance, which is physically impossible, and also that the climate mean state can wander arbitrarily far in any direction, which does not accord with life on Earth”. The first claim, about “an infinitely large variance”, is false; so it will not be discussed further here. The second claim, about how “the climate mean state can wander arbitrarily far in any direction”, is true in principle.
To understand McKitrick's second claim, first note that for “climate mean state” it is enough to consider simply “global temperature”. If the global temperature were truly non-stationary, then it could indeed wander arbitrarily far, up and down; i.e. it could become arbitrarily hot and arbitrarily cold. We know that global temperatures do not vary that much. Hence, global temperatures cannot be non-stationary—this is McKitrick's argument.
McKitrick's argument is easily seen to invalid. Consider a straight line (that is not perfectly horizontal). The straight line goes arbitrarily far up and arbitrarily far down—i.e. arbitrarily far in both directions. A straight line, though, is the basis for the calculations of McKitrick [2014]. Thus, if McKitrick's argument were correct, it would invalidate the basis for McKitrick's own paper.
McKitrick's argument against non-stationarity was raised earlier, by someone else, on the Bishop Hill blog. In response, an anonymous commenter (Nullius in Verba) left a perspicacious comment. The comment is excerpted below.
… everyone agrees that a non-stationary … process is not physically possible for temperature, in exactly the same way as they agree that a non-zero linear trend isn't physically possible. If you extend a non-zero trend forwards or backwards in time far enough, you'll eventually wind up with temperatures below absolute zero in one direction, and temperatures hotter than the sun's core in the other. For the *actual* underlying process to be a linear trend is physically and logically impossible.
However, nobody objects on this basis because everybody knows it is only being used as an approximation that is only considered valid over a short time interval. ….
In exactly the same way, a non-stationary … process is being used as an approximation to a stationary one, and is only considered valid over a short time interval. It arises for exactly the same reason….
Statisticians use non-stationary [models] routinely for variables that are known to be bounded, for very good reason. They're not stupid.
Additionally, McKitrick's argument is an appeal to physics. Yet using physics to exclude a statistical assumption is inherently very dubious. For some elaboration on this, see the Excursus below.
As noted above, several researchers have contended that non-trend-stationarity might be an appropriate assumption for global temperatures. An early paper making that contention is by Woodward & Gray [1995]. That paper currently has 68 citations on Google Scholar, including several since 2013. (One of the latter even presents a physics-based rationale for non-trend-stationarity: Kaufmann et al. [2013].)
There are other papers that do not cite Woodward & Gray, but which also contend for considering non-trend-stationarity; e.g. the paper of Breusch & Vahid [2011]—which is part of the Australian Garnaut Review. Such contending has even appeared in an introductory textbook on time series: Time Series Analysis and Its Applications [Shumway & Stoffer, 2011: Example 2.5; see too set problems 3.33 and 5.3]. Contentions for non-trend-stationarity would not appear in so many respected sources, over so many years, if McKitrick's appeal to a simple physical argument had merit.
It is worth reviewing how McKitrick's story on trend stationarity of the global temperature series changed. First, the abstract of the paper claimed that the temperatures are trend stationary—seemingly an established fact. Second, the body of the paper mentions, in one sentence, that trend stationarity is actually an assumption, rather than a fact—but it gives no justification for the assumption. Third, McKitrick's first e-mail acknowledged that there have been no tests to justify the assumption and also that the validity of the assumption is debated. Fourth, McKitrick's second e-mail, in response to my criticisms of the foregoing, attempted some justifying of the assumption—but with a justification that is easily seen to be invalid, as well as not supported by many other researchers who have studied the issue.
Statistical models
Whenever data is analyzed, we must make some assumptions. In statistics, the assumptions, collectively, are called a “statistical model”. There has been much written about how to select a statistical model—i.e. about how to choose the assumptions.
This issue is noted by the book Principles of Statistical Inference (2006). The book's author is one of the most esteemed statisticians in the U.K., Sir David Cox. The book states this: “How [the] translation from subject-matter problem to statistical model is done is often the most critical part of an analysis”. In other words, choosing the assumptions is often the difficult part of a statistical analysis.
Another book that is relevant is Model Selection [Burnham & Anderson, 2002]. This book currently has about 25 000 citations on Google Scholar—which seems to make it the most-cited statistical research work published during the past quarter century. The book states the following (§8.3).
Statistical inference from a data set, given a model, is well advanced and supported by a very large amount of theory. Theorists and practitioners are routinely employing this theory … in the solution of problems in the applied sciences. The most compelling question is, “what model to use?” Valid inference must usually be based on a good approximating model, but which one?
The book also refers to the question “What is the best model to use?” as the critical issue (§1.2.3).
The selection of a statistical model tends to be especially difficult for time series. Indeed, one of the world's leading specialists in time series, Howell Tong, stated the following, in his book Non-linear Time Series (§5.4).
A fundamental difficulty in statistical analysis is the choice of an appropriate model. This is particularly pronounced in time series analysis.
Note that, in making the statement, Tong does not assume that time series are linear—as the title of his book makes clear.
Concluding remarks
What McKitrick [2014] has done is skip the difficult part of statistical analysis. That is, McKitrick does not genuinely consider the choice of statistical assumptions. Instead, he just picks some assumptions, with negligible justification, and then does calculations.
Realistically, then, McKitrick [2014] does not present a statistical analysis—because the paper is missing a required part. If McKitrick had been forthcoming about this, that would have been fine. For example, suppose McKitrick had included a disclaimer like the following.
The calculations in this work rely on assumptions: about linearity and trend stationarity (and normality). Those assumptions are unjustified and might well be unjustifiable. Relying on different assumptions might well lead to conclusions that are very different from the conclusions of this work. Hence, the conclusions of this work should be regarded as highly tentative.
Such a disclaimer would have been fair and honest. Instead, the paper, especially the abstract, greatly misleads: and McKitrick must have known that it does so.
Finally, methods to detect trends in global temperatures have been studied by the Met Office. A consequence of the study is that “the Met Office does not use a linear trend model to detect changes in global mean temperature” [HL969, Hansard U.K., 2013–2014].
Excursus: Realistic models?
A statistical model does not need to be physically realistic. An example will illustrate this. Suppose that we have a coin. We toss the coin a few times, with the outcome of each toss being either Heads or Tails. We might then make two assumptions. First, the probability of the coin coming up Heads is ½. Second, the result of one toss is unaffected by the other tosses.
The two assumptions comprise our statistical model. The assumptions obviously elide many physical details: they do not tell us what type of coin was used, how long each toss took, the path of the coin through the air, etc. The assumptions, though, should be enough to allow us to analyze the data statistically.
The set of assumptions—i.e. the model—also differs from reality. For instance, our assumption that a coin comes up Heads with probability ½ is only an approximation. In reality, the two sides of a coin are not exactly the same, and so the chances that they come up will not be the same. It might really be, for instance, that the probability that a coin comes up Heads is 0.500001 and the probability that it comes up Tails is 0.499999. Of course, in almost all practical applications, this difference will not matter, and our assumption of a probability of ½ will be fine.
There is also a second way in which our model of a coin toss differs from reality. We can predetermine the outcome of a toss by measuring the position of the coin prior to the toss, measuring the forces exerted on the coin at the start of the toss, and determining the air resistances as the coin was about to go through the air (all this is in principle; in practice, it might not be feasible [Strzalko et al., 2010]). Thus, a real toss is deterministic: it is not random at all. Yet we modelled the outcome of the toss as being random.
This second way in which our model differs from reality—incorporating randomness where the actual process is deterministic—is fundamental. Yet, by modelling the outcome of a coin toss as random, our model is vastly more useful than it would be if we modelled the toss with realistic determinism (i.e. with all the physical forces, etc., that control the outcome of the toss). Indeed, statistics textbooks commonly model a coin toss as being random. Moreover, people have probably been treating a coin toss as random for as long as there have been coins.
To summarize, we model a coin toss as a random event with probability ½, even though we know that the model is physically unrealistic. This exemplifies a maxim of statistics: “all models are wrong, but some are useful”. The maxim seems to be accepted by all statisticians (as well as being intuitively clear). McKitrick, by appealing to a supposed lack of physical realism of non-stationary models, ignores that.
❧ A draft of this Comment was sent to Ross McKitrick; McKitrick acknowledged receipt, but had nothing to say on the technical issues.
See also
| • | Is a line trending upward? |
Breusch T., Vahid F. (2011), “Global temperature trends”, Econometrics and Business Statistics Working Papers (Monash University), 4/11.
Burnham K.P., Anderson D.R. (2002), Model Selection and Multimodel Inference (Springer).
Cox D.R. (2006), Principles of Statistical Inference (Cambridge University Press).
Kaufmann R.K., Kauppi H., Mann M.L., Stock J.H. (2013), “Does temperature contain a stochastic trend: linking statistical results to physical mechanisms”, Climatic Change, 118: 729–743. doi: 10.1007/s10584-012-0683-2.
McKitrick R.R. (2014), “HAC-robust measurement of the duration of a trendless subsample in a global climate time series”, Open Journal of Statistics, 4: 527–535. doi: 10.4236/ojs.2014.47050.
Shumway R.H., Stoffer D.S. (2011), Time Series Analysis and Its Applications (Springer).
Strzalko J.,Grabski J., Stefanski A., Perlikowski P.,Kapitaniak T. (2010), “Understanding coin-tossing”, Mathematical Intelligencer, 32: 54–58. doi: 10.1007/s00283-010-9143-x.
Tong H. (1995), Non-linear Time Series (Oxford University Press).
Woodward W.A., Gray H.L. (1995), “Selecting a model for detecting the presence of a trend”, Journal of Climate, 8: 1929–1937. doi: 10.1175/1520-0442(1995)008<1929:SAMFDT>2.0.CO;2.
Reader Comments (150)
And here are some other statistical correlations that could assume causation - http://www.tylervigen.com/
This unfortunate reality, by the way, applies also to the maths and physics within climate models so those who declare such models are based on the underlying physics and can thus be can be relied upon (even when the fail the essential tests) are either grossly ignorant or out-and-out charlatans.
Sep 30, 2014 at 9:33 AM | JamesG
+1
My comment is alittle bit off. But I do'nt get together the paper of prof. mckitrick et al fromm 2006: Does a Global Temperature exist?, which is denied in that paper und prof. Mckitricks new paper, ehere he uses global temperatures.
To echo what Geoff was talking about, it's amusing to see a battle of wits over "data" that is in effect noise.
Perhaps some appreciation of the raft of assumptions that have been made to produce temperature anomalies is in order before arguing over any trends that appear in it.
Regardless of McK and time-series analysis I don't think the uk.gov (ie the Met Office) or the IPCC means what Doug thinks they mean:-
Uk.gov admits:
"...Global temperatures, along with many other aspects of the climate system, are analysed using physically-based mathematical models, rather than purely statistical models."
But we already knew this! It has been long established in the literature that there is no useful purely statistical model that can be used on the data beyond a basic random walk.
IPCC says:
"The globally averaged combined land and ocean surface temperature data as calculated by a linear trend, show a warming of 0.85 [0.65 to 1.06] °C, over the period 1880 to 2012…."
But they are not saying whether this trend is valid, natural, manmade or a mix: It is just a startpoint for further discussion. All they are saying here is "it's warming!".
The teasing out of any putative manmade from natural warming is done by comparing GCM output (with circular reasoning and now-discredited assumptions about natural variation and aerosols) versus 20th century temps. They have never claimed afaik to use any statistical model beyond simplistic linear trendlines which they and everyone (excepting Tamino and the SkS jokers) know are of dubious, limited value. To this end comparing GCM output with reality is where the real debate is at!
Good point, HelmutU.
"Ross McKitrick has published a paper that deliberately and substantially misleads."
Can you point to the specific references where this is true and highlight what differences the 'misleading' assumptions makes. Ross always comes across as pretty open about his work, I just find your references not quite specific enough to understand your point.
Doug,
Your accusation that I have “deliberately and substantially mislead” readers is completely out of line. You take issue with the fact that I made an assertion about trend stationarity, not because you have evidence to the contrary but because you assume the contrary. If you look at my online code you will see the line for the unit root testing. Since the data are known to have at least one trend break I use the Zivot-Andrews test for a null of a unit root. The 1% critical value for the ZA test is -5.57. The ZA test statistics for Hadley, UAH and RSS are, respectively, -862.6, -874.1 and -885.6. I didn’t report them in the paper but they massively reject the unit root null. The implied alternative, as I did state in the paper, is trend stationarity. If you want to set up and estimate a model explaining why you think the data do actually have a unit root, be my guest, but the fact that you would publicly accuse me of misleading readers merely on the basis of your assumption that the data have a different property makes it clear that further attempt to discuss this with you is pointless.
Bishop, can I request that you remove Doug's accusation against me of misleading readers, as it is false and derogatory.
Doug, Dealing with your other responses:
(1) You say "If you have no idea what a nonlinear time series is then perhaps you should consult books on the subject." What I said was I have no idea what you meant by saying I assume all the relevant equations for the noise are linear. BECAUSE THE HAC NOISE MODEL IS NON-PARAMETRIC AND THE “RELEVANT” EQUATIONS ARE NOT WRITTEN DOWN. That’s the whole point of the HAC model, it is general to any form except a unit root. An assumption of linearity of the noise equations is not imposed. I said this quite clearly in my reply, but you distorted it to make a ridiculous insult for a response.
(2) As I made clear, I make no assumptions about linearity of the global temperature series, in fact by modeling a pause, it implies a nonlinearity in the trend. I estimate a linear trend term over a sub-interval. You might find that uninteresting because you would like to see a nonlinear trend estimated over that interval. Fine, get the data and estimate it. My calculations were for the benefit of people who want to know what is the 95% confidence interval around a linear trend term over the stated intervals. If you don’t find that an interesting piece of information you are free to ignore it and write papers explaining what you think would be interesting.
(3) The estimator of the I(1) variance is finite in a finite sample, and grows monotonically as the sample grows, which itself is an unphysical characteristic even in finite time. More to the point, what matters is the underlying moment itself. The long run variance itself is infinity, irrespective of the sample size.
(4) I am computing the interval over which the trend is statistically indistinguishable from zero. I am not extrapolating the line infinitely far in either direction. That’s why I report the start and end dates, and draw conclusions based on the interval they contain.
(5) A “stochastic trend” is another term for the drift component of a unit root. By saying the data are trend stationary I am saying they do not contain a unit root. If they contain a unit root it is not valid to report a confidence interval around a linear trend.
Doug,
Your premise is right on:
All parties need to be held to high standards.
Your execution isa wee bit questionable.
You are not discussing the differences of opinion you are having with McKttirck in a reasonable or frankly rational manner.
You also seem top be confusing your opinion with fact, since what Ross is answering to you seems to be quite reasonable.
Additionally, you seem to be obsessed on a minor point and are completely willing to trash his character, motive, and professionalism over somethign that frankly appears trivial.
This is too bad, since your work in deconstructing the IPCC seems to have been absed on substantive issues that the IPCC promoters used to mislead a whole generation of policy makers.
You might want to catch your breath a moment or two, and consider if what you are doing has anything to do with actually moving the issue forward or even holding Ross to high standards.
So far you seem quite unconvincing.
Respectfully,
etc
In citing Ross’ first, second and third emails to him as evidence of Ross knowing “trend stationarity is nowhere near being an established fact”, the emails appear to be addressing stationarity of global temperatures, not their trend stationarity, which is a different thing. Therefore that section of Douglas’ case is irrelevant.
To make a stronger argument, Douglas could have provided results of statistical tests such as the KPSS test that indicate the absence of trend stationarity, but he didn’t, and from my back of the envelope probably can’t.
On the five bullets between Ross and Douglas in the comments, I score the result 5-0 in Ross’ favour.
Finally, I believe Douglas has failed to justify his use of extreme rhetoric, and I support Ross’ request for the removal of the accusation of misleading readers.
As usual I find myself agreeing with hunter (how does he do that?). In strictly formal terms Douglas' criticisms may well be valid but they appear irrelevant to Ross estimating a linear trend over a sub-interval.
Massive thanks and respect to Douglas for what he has done re the IPCC but to have launched such a personal attack on Ross in this instance is clearly unnecessary. I hope it's not to late for both parties to stop to draw breath.
is this another occasion where Mr Keenan ups the rhetoric and accuses academics of fraud....then has to back down, when it is shown that he is being overly aggressive in his rhetoric? Remember his rant about carbon dating that got debunked so effectively by Radford Neal?
Agree with the above commenters, Keenan has thoroughly disqualified himself with the ridiculous personal attacks at Ross McKitrick. I fail to understand why hte Bishop did not removed the attacks before publishing.
This started with an unpleasant air and has turned into an exercise in gratuitous nastiness on Doug Keenan's part - taking Ross McKitrick's reply, twisting it into something different and then ridiculing McKitrick for the reply he did not make.
Bishop Hill - I'm for deletion too.
(haven't read any of it, and will not waste any of my time even trying to, but ...)
we should not try to replace complete rubbish with (what we think is) better rubbish ...
leave the rubbish to them ...
whatever we say, whatever we try to discuss with them, whatever we try to "calculate", it will cost us anyway ...
but time will vindicate "sceptics" ... because their theory, and the data they fabricate/use are utter bollocks ...
truth (and science) will prevail in the end ... as it always has, as it alway will ...
Jonathan,
After re-reading my dyslexic post, all I can say is thanks for wading through and seeing my point despite my poor typing.
Sep 30, 2014 at 6:31 PM | Martin A
Totally agree (even though I do understand NiVs point earlier). This post and thread tells you more about Doug himself than anything else.
When Doug says
The calculations of McKitrick [2014] rely on certain assumptions. In principle, that is fine: some assumptions must always be made, when attempting to analyze data. A vital question is almost always this: what assumptions should be relied upon? The question is vital because the conclusions of the data analysis commonly depend upon the assumptions. That is, the conclusions of the analysis can vary greatly, depending upon the assumptions.
It confirms everything I have believed about statistics. They can be used to prove anything!
I operate under a very simple precept - if something needs to be proved by complicated statistics, it probably does not exist at all.
Perhaps the Bishop should consider jsut deleting most of Doug's inflammatory knit picking?
hunter: "Perhaps the Bishop should consider jsut deleting most of Doug's inflammatory knit picking?"
Yes, DK needs to be protected from himself. Not sure why he felt it necessary to be so aggressive and insulting towards RMcK.
Regardless of the merits of DK's critique of the paper, a public apology for the personal attack on a professional colleague is called for.
TC: I agree completely. Why is Keenan so angry?
The assumption of a linear trend will eventually take temperature unreasonably high or unreasonably low. However, when attempting to pinpoint the beginning and end of a pause, the trend is zero. A zero linear trend can be trend stationary forever without reaching an unreasonable temperature! As for the trend at the beginning and end of a pause, that trend doesn't need to persist indefinitely long - and thereby produce and unreasonable temperature. Those trends only need to last longer enough to define a beginning and end for the pause.
Climate systems may be chaotic and non-linear, but that doesn't mean that all of the observables in such a system must exhibit non-linear behavior. The IPCC is perfectly happy, for example, defining climate sensitivity - to a first approximation - as a linear relationship between forcing and temperature. Well-behaved functions can always be approximated by a linear function (two terms from the Taylor series), with the important question being how far can one go before the error becomes to large. Temperature change arises from integration of radiative imbalance (and heat capacity per unit area). Even if a radiative imbalance were discontinuous, the resulting temperature change would be continuous.
In the 1990's, a relatively constant radiative imbalance (say 1 W/m2) could have caused a roughly linear increase in temperature. The end of the 1997/8 El Nino might have "permanently" increased upwelling off of the South American coast, and changed that average radiative imbalance to zero and created the pause.
As a lurker whose decades old math 'expertise' is totally inadequate to deal with the substantive issues, I do have a question:
If Keenan had issues with McKitrick's work, did he or did he not approach McKitrick privately before initiating a distasteful public pi$$ing contest?
Doug's comments should not be censored. The risk from censorship is too high.
The discussion is valid except where misquotations happen or wrong inferences are drawn.
The discussion is mainly about the weight to be given to extreme formalism about our present ways of expression of the statistical basis.
There is value in Doug reminding readers of the need to comprehend the formalism, especially when it affects a claimed outcome.
There is no evidence that Ross does not comprehend the formalism, which in any case is in continuous improvement and not necessarily correct at any point in time.
Doug has not shown that Ross made errors of any significance by choosing to operate a little informally, as most authors do when their stats are applied to climate analysis.
Right or wrong, the discussion reminds me of what happens to the constant that is in the formal expression of the indefinite integral. Useful results are still possible informally, by ignoring it and by refraining from repetitive purism discussions.
Overall, for practical purposes, insignificant harm results from the informality of the Ross approach. Most reasonable people would agree that, working back from today, using lists of supplied numbers, the numbers show little variation until we reach back a certain number of years, that number depending inter alia on the list chosen, such as satellite soundings versus thermometric reconstructions
(My apologies for mention of the Breusch paper in a post above, when it was already in supplied in the original list of references for the article.).
The only intentional misleading that's being done that's perfectly obvious to me here is Keenan's intellectually vacuous and intellectually dishonest misquotation, as if McKitrick had said he had no idea what a nonlinear time series was. This would seem to be pretty compelling evidence of Keenan's psychological projection of his own shortcomings onto McKitrick.
Doug, if you can't be honest in small matters, why should we believe you in large ones?
@ Geoff Sherrington. "Doug's comments should not be censored. The risk from censorship is too high."
The Bish now and then (and rightly) deletes stuff that is rude or over the top or simply off topic.
There is a difference between censoring opinion or comment on technical matters and censoring outright nastiness.
@ Ross McKitrick
I am checking some references and will come back on this shortly. One reference that I am checking is Vogelsang & Franses [2005]. I have a question related to that, which I hoped you would answer.
V&F present a particular test for time series. They say that, for their test to be applicable, the series must satisfy a certain technical condition (§2.1). They then say that the condition “rules out stationary time series with long memory”.
Your paper applies their test to the global temperature series. Yet the global temperature series has been argued, by many people, to have long memory. What is the justification for applying their test to the temperature series?
Technically, Doug has not misquoted. He accurately quotes three words from Ross: "have no idea". But Doug does appear to mischaracterize what Ross says he has no idea about. Now, it's not obvious that, in so mischaracterizing, Doug does so intentionally. Just as it's not obvious - as Doug claims - that Ross "deliberately...misleads" in his paper.
I just wish folk would be a lot more circumspect about levelling accusations of intentionality. It really does poison the atmosphere.
What a stushie! I am late to it, but my first impression is that Ross McKitrick has made an analysis based, as Doug Keenan notes, on stated assumptions. So far, so good. But DK does not think the assumptions are justified and seems to be annoyed that RM does not agree and does not respond quickly enough to his emails. Then DK goes over the top, and what should have been a routine, reasonable, and rational discussion threatens to be derailed. RM remains courteous, admirably so, and more exchanges take place.
Now, I need to find time to look at the RM paper, and the DK remarks, and see if I can manage to clarify things further at least for myself. But in the meantime, it seems clear that DK owes RM an apology for poor manners.
Mr Keenan must be fuming all the time. Just pick up any journal and you'll see statistical misuse and abuse all over the place. Besides McKitrick is dealing with a finite series (1980 onwards?) so his statistical model is at least sound for the given series assuming the series is trend stationary. The cited comment referring to extrapolation of trend (to below absolute zero) could be equally applied to spatial examples such as gradient slope on a hill (topography is also non-linear) - the point seemed a little contrived. Are we saying gradients on slopes are statistically invalid and therefore by extension physically meaningless or misleading; in terms of mechanics they are completely valid. Nearly all stress tensors defined slope stability prediction are dependent on these gradients.
I think this is an awfully pedantic especially when one considers what passes peer review. Also consider that the big brains in hedge-funds often rely on extrapolated non-parametric and parametric models that assume linearity - for time series. In short, statisticians need to calm down, near-enough is often good enough even if the outcome is biased because of poorly defined methodologies.
John Shade
I think DK has been quite polite but to the point as has RK. My own experience is that statisticians and mathematicians give the most instructive and polite reviews. The scientific reviewers are often pompous and full of self-importance.
DK is obviously a very bright man but I think his pet-hate is in danger of becoming an obsession. Perhaps the answer is that all scientific papers must have one statistician/mathematician as a reviewer and two scientific reviewers.
I'm with the Keenan guy, if not but just for the sheer amusement with this kerfuffle
Anyway without context any time series is irrelevant IMHO
You cannot draw conclusions from any standalone time series.. Any
Thanks to readers who have weighed in on the question of whether I should delete Doug's more inflammatory postings (or not). I originally suggested to Doug that he remove these sections from the original posting. He pointed out that he would say the same thing about someone on the alarmist side of the debate and pointed out, not unreasonably, that the article was already posted at his own website.
That being the case, I'm not sure what I achieve by removing the offending passages here. I'd welcome further thoughts from readers here though.
Leave it up, your honor.
It is a good lesson that even well established skeptics like Mr. Keenan have their limits.
As Dirty Harry would point out, "A man's got to know his limits".
Hopefully Mr. Keenan will learn he has limits and to respect them.
Leave it up, if only to prevent the 'other side' from claiming you are whitewashing on behalf of Ross. Unless totally devoid of factual content, what has been said should be left still said.
RichieRich, I'm generally in favor of circumspection as to claiming you know sombebody else's heart.
But Keenan made a decision where to turn those quote marks on an off. The obvious explanation for the start and stop of quotations at "have no idea" is for Keenan to make McKitrick seem lazy and foolish; the quote could have just continued!
The more circumspect explanation-- that Keenan maybe was just too furious to be able to read intelligently past "have no idea" to what was actually being said after-- not only seems implausible, but is no less disqualifying as to whether to take someone's discourse / invective seriously than seeing it as intellectual vacuity or dishonesty. If you can't fairly read, you can't fairly criticize another for being misleading.
This sort of behavior really ought to matter.
As I said earlier, I don't think we should treat fellow sceptics any less critically than warmists - it would be hypocritical to do so and then claim we were on the side of science. I'm also against censorship and editing the record of events generally. And of course it would mean that every warmist scientist who thought posts or comments here were "false and derogatory" could turn up and demand the post be removed.
It would be doubly hypocritical to extend that facility to our friends and not our opponents. To turn this into the sort of echo chamber you see on the other side of the debate where opposition is not tolerated and comments they don't like are deleted. Sceptics have been pretty vocal about places like RealClimate doing it - do you want to open BH up to the same criticism? (I have in the past defended BH against that accusation - it's personally disappointing to me to find out that I was wrong and they were right.)
However, it would be fair enough (although in my view not strictly necessary) to add an update or caveat to the top post to say that Ross replies below and disputes the criticism. That way, anyone who would ordinarily not read all the comments would be made aware that the claims were in question.
Tell people the complete truth, give people the argument from *both* sides and let them make their own minds up. Freedom of speech, and freedom to take the consequences of what you say.
NiV +1
NIV - whatever I might have said earlier in this thread, I'm against having comments deleted because they challenge opinions or statements on scientific matters. On the other hand, as a previous commenter said, it poisons the atmosphere to accuse others of deception and so on.
I think the Climate Audit has a good policy - anything goes so far as contradicting views on statistical or scientific matters is concerned. However ascribing malign motives to others and particularly accusations of fraud are not tolerated at all.
Just to be clear, I am not advocating that Doug's post be deleted--I am more than happy for people to read it and my responses. I am requesting that his accusation that I "deliberately and substantially" mislead readers be removed since it is false and harmful. It is not part of intelligent or respectful debate, it is mere slander.
"I'm against having comments deleted because they challenge opinions or statements on scientific matters. On the other hand, as a previous commenter said, it poisons the atmosphere to accuse others of deception and so on."
I tend to be a bit of a purist when it comes to free speech. As soon as people start saying "But free speech doesn't include the right to say..." and then listing all the forms of speech they don't like and would like to ban, that's not free speech, and the people saying such things I consider to be opponents of it.
Free speech does include the right to be offensive, rude, to be factually wrong, to lie, and to advocate for the immoral. That can sometimes be unpleasant, but it doesn't actually hurt unless you let it. "Sticks and stones may break my bones, but names will never harm me" as the nursery rhyme goes. You can ignore it, avoid it, complain about it, or return fire. But when you start trying to ban it, forbid it, or punish it, you're on a slippery slope.
Worse, political correctness has a morally corrosive effect because it effectively forces people to tell lies - to hide their true opinions, and to become deceivers themselves. You have to constantly censor your own thoughts and words, so you don't accidentally say the wrong thing, which normalises the principle of censorship generally. You avoid criticising colleagues, or bury criticism in mealy-mouthed diplomatic doublespeak, which means tolerating their failures - as other scientists have with the failures of climate science. Errors fester in the darkness. Lies multiply.
That said, on somebody else's private property, the owner gets to set the rules. So if Andrew decides it's being disruptive he can enforce his own comments policy. And of course, he will be judged on that himself. I've always admired his commitment to free speech. It would be a shame, I think, if he had to change that.
The other difficulty with constraints on free speech is that it is very difficult to be consistent. People are routinely derogatory about others here - scan a few threads to see what it said about climate scientists, environmentalists, eco-campaigners, climate journalists, green politicians, etc. Quite a lot of them have been accused of deception, too. Should Andrew tolerate that?
Bearing in mind that he's complained about it numerous times, and asked people not to do it, he clearly doesn't like it. Should Andrew delete Doug's comment for being derogatory, but then go through every thread and also delete every single comment that is derogatory or accusatory about a climate scientist or green? Because if he does, a lot of the threads are going to start to look pretty bare.
And it also seems to me to be pretty odd to me for participants in the climate war to be so sensitive. There are lots of people who accuse climate sceptics of being deceptive, stupid, or of being outright paid liars. There are lots of people who come on here to tell us we're wrong/deceptive. And we go on other people's sites and try to leave comments telling them that they're wrong or being deceptive, very often in quite pungent terms. Are you happy for them to delete your comments if they don't like them? Does it bother you that someone like ATTP or Willard or BBD can come on here and say we're derogatory deceivers and have their comments left alone, but Doug can't?
I think Doug's comments touched a nerve with many here, because we're supposed to be "on the same side". He, on the other hand, is trying to adhere to a principle: that we should be as critical with our own side as we are with the opposition. He's said the same sort of things about the Met Office and the IPCC, and we all cheered him on. Personally, I'd be happier if we were all a bit more cautious and polite about criticising both sides, but I respect someone willing to stand up for the principle of impartiality even more. Perhaps it ought to lead the rest of us to give a little more thought to how we speak about the warmists.
I don't know. I hope Doug will learn more caution about adding in the appropriate caveats into what he says from this. However, I believe he has tried to be honest, and said what he thought was the case based on what he knew and understood. I hope he holds to his principles of saying it as he sees it, and doing so for all sides impartially. If people don't like that, they can ignore him, or complain. And I hope he continues to pursue the technical question here - I have to say, I don't myself understand how some of Ross's claims can be right, but I haven't examined his arguments and those in the supporting papers in enough detail to be able to say either way. It's a shame when high emotions get in the way of the science.
Bish: It's your blog, You can remove anything you feel is inappropriate (or worse exposes you to legal liability). I would prefer that you leave as much of the post intact as possible to avoid censorship, but you could delete a passage or phrase and perhaps link Doug's site or the WayBack machine for any passages that you change.
Doug: I doubt you can demonstrate that Ross intended to mislead readers when he published his paper. If Ross doesn't respond to your criticism, inaction by Ross certainly becomes deliberate at some point, but you are still debating whether readers will be so "substantially" mislead that inaction becomes problematic. Ross's ABSTRACT states that:
"HAC-robust trend variance estimator which is valid as long as the underlying series is trend stationary, which is the case for the data used herein". Not all temperature data is trend stationary, but readers have been clearly warned about this issue.
Ross and Doug: Another question concerns whether a linear model is a suitable model - and perhaps the only sensible model - for identifying a "pause". What if I wanted to study the incidence of "pauses" in output created by a random walk (or other process)? Is this a question that could be addressed by Monte Carlo methods? How many pauses can Ross's method find in random walk output before the method is invalid? (The answer obviously is linked to alpha.) If one finds too many pauses, how much too long is the calculated pause. (Shorter pauses will certainly be found more often by chance.
of course Ross's wrong.
There is no way to draw any conclusions from time series on climate, because we do not understand the system well and have no context except for 150y of very dodgy records and virtually NO records on proxies etc before that.
the first thing youwould need is a couple of earths besides each other and compare the results over a few million years with varying conditions between them.
Time series, if you have 150 points in the series and each point can take say 100 values, you have 100**150 possible time series..
particles in the universe blahblahblah..
Then some warmish parasite tells you , but temp should be like it was 2-3 centuries before the records began.. what he really means is he wants to have 100**500 time series for the discussion to be spinned appropriately for his wallet
particles in the multiverse blahblahblah
If you would want to say such and such time series has a tendency to , say, rise at an end, you would have to justify that by comparing it with some others under slightly other conditions,
Only one attempt I have seen so far to compare earth with venus in an interestig way was this chap HarryDaleHufmann
Anyway
cautious talk polite mellow blahblahblah
=> I like hardtalk
a steak has to be flamed on both sides to be pallatable.
Oct 2, 2014 at 7:45 PM | Nullius in Verba
From the Stanford Encyclopedia of Philosphy's article on Freedom of Speech
I very much enjoyed your post. But I'm guessing that, though you class yourself as something of a purist re free speech, you would nevertheless support certain limits being placed upon speech.
Not that any of this directly addresses the particular issue of whether sections of Doug's posts should be deleted. My own view is that, on balance, they shouldn't - but, at the same time, I'd like to see a great deal more circumspection with regard to making statements about people's intent.
"But I'm guessing that, though you class yourself as something of a purist re free speech, you would nevertheless support certain limits being placed upon speech."
Yes. The standard libertarian approach here is the harm principle, as defined by JS Mill. Libel and slander occurs when the speech has direct material or financial consequences. (E.g. loss of business.) Child pornography is not itself the issue, but doing harm to children to produce it is. (The distinction matters. There's an interesting ethical issue here with regard to cartoons, and computer-generated graphics. Many libertarians would argue they do no harm.) Advertising is allowed - I'm guessing you meant false advertising. If it causes people to take a financial loss, when they buy something that doesn't work, that financial loss is a harm. And state secrets are secret precisely on the basis of the material harm that their dissemination might do.
In all these cases, it is not the speech itself that is the issue, it is the material consequences of that speech, and the harm done by it. It is not simply a loss of reputation, it is the direct harm done as a result of that loss of reputation. If somebody does you harm unjustly and without your informed consent - whether by punching you in the nose, or stealing your possessions, or by ordering either of those done - you can claim recompense for that harm. But mere bruised feelings is not a significant enough harm to invoke such severe measures. The harm done by stifling free speech outweighs it.
Nor can you anticipate it. You cannot deprive somebody of their rights in case they might do you a harm. You can only make a claim against them when the harm has been done, or is otherwise unavoidable.
The principle is that we should allow the maximum freedom to everybody consistent with not impinging on the freedom of others. As you correctly point out, that doesn't mean absolute unconditional freedom. It cannot.
For example, and as I already said - the ownership of a venue trumps the 'free speech' rights of any visitors. Clearly that's not absolute free speech.
That's not to say it is ever a good thing - just that it might sometimes be the lesser of the two evils.
Actually Ross is right, ptw. He clearly frames the terms and conditions for his written study in the abstract:
"Ross's ABSTRACT states that:
"HAC-robust trend variance estimator which is valid as long as the underlying series is trend stationary, which is the case for the data used herein"." (h/t Frank)
I'm reminded of the mule training 2X4; gentlemen, you have my attention.
===================
Doug Keenan,
I think that you are technically correct on the main issue - trend-stationarity, and will post a note for Ross on the subject. However, IMO you have managed to muddy the waters very seriously here by (a) imputing malicious intent and (b) introducing some red herrings which on the face of it are either too esoteric to understand or just plain wrong. The fact that we may be dealing with a "coupled non-linear chaotic" system does not ipso facto imply that the temperature series cannot be trend-stationary, and is barely relevant to your main point. And, like Ross, I have no idea what you mean by "the paper also assumes that all the relevant equations (for the noise) should be linear".
I would strongly suggest for your own effectiveness (and peace of mind) that you consider a change in rhetoric, and more profoundly, that you maintain the view for as long as you possibly can that someone who disagrees with your perspective might still be arguing in good faith.
Professor McKittrick,
Like Keenan, I end up thinking that your paper is misleading - although I abhor the unnecessary accusatory tone that he has used to describe his concerns, and I am most definitely not assuming any ill intention on your part.
(1) Every analysis of the longer annual-averaged GAT series suggests that we cannot reject a unit root in the series. Ergo, It is not trend-stationary over the full series.
(2) Spectral or empirical mode decomposition of (any of) the longer annual-averaged GAT series suggests that the series can be partitioned into (a) a smoothly varying, monotonic, slightly non-linear secular series, (b) a remarkably regular multidecadal oscillatory cycle and (c) a high frequency red noise series, associated with ENSO and volcanic events. Any trend line fitted over the length of the full series shows significant cyclic heteroscedasticity – directly associated with the multidecadal oscillations.
(3) If the temp series is modified to test for a unit root after subtraction of the multidecadal oscillatory cycles, the residual series, comprising the sum of the secular series plus the high frequency red noise, shows no evidence of a unit root. The implication is clear – the evidence of long-term persistence comes only from the multidecadal oscillations. In the GAT, these dominant oscillations resolve into periods of around 22 years and a bit over 60 years.
The long-term series simply is not trend-stationary under any rigorous testing , (as I think you agree).
By examining short-enough segments of the full series (or the satellite series which only represents about a half-cycle of the dominant quasi-60-year cycle) it is possible to “justify” alternative structural models, but this does not make such models correct or useful - even over those same sub-intervals of the series.
Consider the argument put forward by the Met Office that the warming over the 20th century was “statistically significant”. Since the implied underlying statistical model was completely spurious, I am sure you would agree that the statement was misleading. Suppose however that the Met office had considered later and later dates until they found a year X where a test on the forward series segment rejected a unit root in the data and failed to reject trend-stationarity. If I follow your logic correctly, since the Met Office could then argue that the data was trend-stationary over the sub-interval, then it would then be legitimate for the Met Office to report that, at the least, the warming since year X was statistically significant.
Your paper, coming as it does from an individual with a highly respected appreciation of statistics, IMO will propagate the inappropriate use of statistical inference from trendlines drawn willy-nilly on sub-intervals of series when the full series does not have the structural properties to justify such inferences.
The abstract for McKitrick [2014] states that the method used is “valid as long as the underlying series is trend stationary, which is the case for the data used herein” (emphasis added). Anyone reading that would reasonably conclude that the data “is” trend stationary. In fact, trend stationarity is debated by researchers, as even McKitrick has acknowledged.
Moreover, trend stationarity is a vital issue for the paper: because without trend stationarity, the conclusions of the paper are wholly unfounded. It is not up to me to show that the data are not trend stationarity, as a few people illogically claimed. It is up to McKitrick to demonstrate trend stationarity—or, alternatively, say that we do not know if the data is trend stationary. This should be clear to everyone.
Thus, as my post put it, “the paper, especially the abstract, greatly misleads: and McKitrick must have known that it does so”. Or, as one of my comments put it, the paper “deliberately and substantially misleads”. Some people do not seem to want to accept that. A few people have even gone farther, saying that I should not be allowed to speak—i.e. not be allowed to point out that a prominent skeptic has done such a thing. Such people are copying some of the worst traits of global-warming alarmists: refusing to face unpalatable evidence, and trying to silence those who bring the evidence forward.
Martin A (Oct 2, at 10:39 AM) asserts that Bishop Hill should be more like Climate Audit where, he claims, “ascribing malign motives to others and particularly accusations of fraud are not tolerated at all”. Compare the claim with reality. For example, there is the Climate Audit post “Did Jones et al 1990 “fabricate” its quality control claims?”, which is about my allegation of fraud against Wei-Chyung Wang. Another example, from a couple days ago, is the post “Sliming by Stokes”, which states that an article by Nick Stokes contains “a series lies and fantasies”. There are many many more examples.
People who want different standards for criticism of alarmists and criticism of skeptics should consider what their motives are. Tribalism? A belief that they are on the right side, and so they should be allowed to get away with more? A desire to believe that there is only goodness in any person who is a skeptic? Whatever the case, the same excuses will surely have been used on the alarmist side. Consider that.
Ross McKitrick stated, in a comment (Sep 29, at 6:49 PM), that he “had no idea” what I meant by saying he assumed all the relevant equations for the noise are linear. I misinterpreted what he stated, and some people have criticized me for doing so. I understand the criticism. My misinterpretation, however, was to McKitrick’s benefit. What turns out to be the correct interpretation—as explicated by Ross McKitrick in a later comment (Sep 30, at 4:11 PM)—is worse.
The later comment says the correct interpretation is that the “NOISE MODEL IS NON-PARAMETRIC AND THE “RELEVANT” EQUATIONS ARE NOT WRITTEN DOWN” (caps in original). But that is impossible, and anyone with a basic understanding of models of time series would know so. Could it really be true that McKitrick does not know so? I did not believe that, at first.
Matt Briggs, who is a statistician, has a blog post about McKitrick’s paper. The primary point of the post is that McKitrick’s paper relies on a statistical model and yet there is nowhere near adequate justification for that model. (That is one of the main points in my post as well, of course.) McKitrick left a comment disputing some of Briggs’ post. Briggs replied, “I don’t think you [Ross McKitrick] … really grasp what statistical models are”. He followed that up with a post entitled “The true meaning of statistical models”.
Statistical models are the basis of statistical analysis. In statistics, inferences are not drawn directly from data; rather, a statistical model is fit to the data, and then inferences are drawn from the model. Anyone who does not grasp what a statistical model is must be severely statistically incompetent. I misinterpreted McKitrick’s statement because I did not conceive that he could be so incompetent.
The story gets worse. The paper of Vogelsang & Franses [2005], cited by McKitrick [2014], presents a statistical method for comparing trends in two series, under certain conditions. McKitrick [2014] essentially just applies the method to the global temperature series—that is all McKitrick really does.
Vogelsang & Franses base their method on a statistical model—as they must. The first three pages of their paper are on my web site (only the first three, because of copyright restrictions). Page 3 describes the statistical model. On that page, the equations are indeed written down—as they must be. Yet, as noted above, McKitrick claims the “EQUATIONS ARE NOT WRITTEN DOWN”. McKitrick’s claim is false, and McKitrick surely knows that it is false.
There is another issue arising from Vogelsang & Franses. V&F explain that one of their model equations “rules out stationary time series with long memory”. Yet many researchers have argued that the series of global temperatures has long memory (also called “long-term persistence”). Indeed, McKitrick has even published on this topic. McKitrick [2014] acknowledges some of that, saying “there is empirical evidence that the error processes exhibit … long term persistence”. Thus, McKitrick acknowledges that long-term persistence is evidenced in the data, and yet he relies on a statistical method that he knows does not work with data that has long-term persistence. Is this not fraud?
Finally, McKitrick’s other replies are also rhetorical. For example, consider point #4. Point #4 began when McKitrick e-mailed me to say this: “The reason I do not accept the nonstationarity model for temperature is that it implies an infinitely large variance, which is physically impossible, and also that the climate mean state can wander arbitrarily far in any direction, which does not accord with life on Earth”. McKitrick’s argument is vital for McKitrick’s paper: if temperatures are modeled as nonstationary (and not trend stationary), then the paper’s calculations are invalidated.
My post treats McKitrick’s argument at length. McKitrick and I have also discussed the argument in comments here. Relevant extracts from those comments are below.
If a straight line has start and end dates, then indeed its length is finite. If there are start and end dates, though, then the variance is plainly finite too; so McKitrick’s argument for not accepting a nonstationary model is then voided. This is obvious; i.e. this is surely known to McKitrick.