There’s a saying, “timing is everything”. After reading this, I think it is more true than ever. In other news. Paul Vaughn is giving Bob Tisdale serious competition in the contest over who can fit the most graphs into a single blog post. ☺ There is a helpful glossary of symbols and abbreviation at the end of this post that readers would benefit from reading before the essay. A PDF version is also available via link at the end of the article. – Anthony
Guest post by Paul L. Vaughan, M.Sc.
Lack of widespread awareness of the spatiotemporal nature of interannual terrestrial oscillations is perhaps the most paralyzing bottleneck in the climate discussion.
North Pacific Pivot
Elegant factor analyses by Trenberth, Stepaniak, & Smith (2005) concisely chart the limits of linear climate exploration, providing strong clues that the North Pacific is a globally pivotal intersection.
D – T = -SOI (an index of El Nino / La Nina – details below in “Data & Symbols” section)
WUWT readers are well-acquainted with Tsonis, Swanson, & Kravtsov (2007). Recently Wyatt, Kravtsov, & Tsonis (2011b) shared the following on Dr. Pielke Senior’s blog:
“PNA participates in all synchronizations.”
Orientation for ENSO- & PDO-centric readers:
Complex Correlation
Simple linear correlation can do a part-way decent job of summarizing the preceding intrabasin relations, but properties of interbasin & interhemispheric multiscale spatiotemporal relations clarify the need for complex summaries. For example:
Limitations of linear methods are emphasized by Maraun & Kurths (2005). A mainstream audience might not appreciate their beautifully concise section 3 primer, but there’s a simple way to look at interannual spatiotemporal phasing.
The ~2.37 year signal which is so prominent in the equatorial stratosphere is also easily detected in the troposphere, but there’s clearly “something else” contributing to interannual tropospheric variation.
Note that when iNPI’ doesn’t “go with” iAAM & iLOD, it “goes against” them, much like a switch that is either “off” or “on”. Specialists like Maraun & Kurths might speak of coherence and illustrate the nonrandom distribution of phase differences.
Multiscale complex correlation (for example using adjacent derivative based complex empirical wavelet embeddings) can measure complex nonstationary relations where simple linear correlation fails catastrophically. Naive investigators unknowingly encounter Simpson’s Paradox by falsely assuming independence and blindly running linear factor analyses (such as PCA, EOF, & SSA) without performing the right diagnostics.
Northern Hemisphere Inter-Basin Interannual Coherence
Nonrandom phase relations explored by Schwing, Jiang, & Mendelssohn (2003):
Interannual Solar-Terrestrial Phase-Relations
Terrestrial phase relations with interannual [not to be confused with decadal] rates of change of solar variables, including solar wind speed (iV’), are nonrandom:
Inter-Hemispheric Interannual Phase-Relations
For those wondering how AAO & SAM fit in:
Global Synchronicity
Synchronicity’s the norm. Orientation, configuration, amplitude, & extent of globally constrained & coupled jets & gyres are pressured while network monitoring remains stationary. Regional temporal phase summaries are intermittently flipped by the stationary spatial geometry of monitoring networks in the turbulent global context.
Note particularly (in the last 2 graphs) the strong & stable interannual synchronicity of northern annular, southern annular, & global modes for the decade beginning ~1988. The commencement of the pattern coincides with concurrent abrupt changes in Arctic ice flow (e.g. Rigor & Wallace (2004) Figure 3) and European temperature (e.g. Courtillot (2010)).
Local Connection
Here’s how NPI relates to minimum temperatures at my local weather station:
Concluding Speculation
Terrestrial geostrophic balance is affected by the concert of changes in:
a) interannual (not to be confused with decadal) solar variations.
b) decadal amplitude of semi-annual Earth rotation variations – [see Vaughan (2011) & links therein].
c) solar cycle length – [see links in Vaughan (2011)].
Nipping Potential Misunderstandings in the Bud
“So you’re claiming the North Pacific controls global climate?”
No.
Why do I hear the same places mentioned every rush hour on the traffic report? Bottlenecks are easy places to detect changes in pressure & flow (whether global &/or locally intersecting), even using the simplest methods. Methods such as those suggested by Schwing, Jiang, & Mendelssohn (2003); Maraun & Kurths (2005); and Tsonis, Swanson, & Kravtsov (2007) help expand our vision towards the rest of the network. We have a lot of work to do (both exploratory & methodological).
Further Reading
Everything written by Tomas Milanovic at Dr. Judith Curry’s blog Climate Etc.
Vaughan, P.L. (2011). Solar, terrestrial, & lunisolar components of rate of change of length of day.
Referenced Above
Courtillot, V. (Dec. 2010). YouTube Video (~30min): Berlin Conference Presentation.
http://www.youtube.com/watch?v=IG_7zK8ODGA
Maraun, D.; & Kurths, J. (2005). Epochs of phase coherence between El Nino-Southern Oscillation and Indian monsoon. Geophysical Research Letters 32, L15709. doi10.1029-2005GL023225.
http://www.cru.uea.ac.uk/~douglas/papers/maraun05a.pdf
Rigor, I.; & Wallace, J.M. (2004). Variations in the age of Arctic sea-ice and summer sea-ice extent. Geophysical Research Letters 31. doi: 10.1029/2004GL019492.
http://iabp.apl.washington.edu/research_seaiceageextent.html
Schwing, F.B.; Jiang, J.; & Mendelssohn, R. (2003). Coherency of multi-scale abrupt changes between the NAO, NPI, and PDO. Geophysical Research Letters 30(7), 1406. doi:10.1029/2002GL016535.
Trenberth, K.E.; Stepaniak, D.P.; & Smith, L. (2005). Interannual variability of patterns of atmospheric mass distribution. Journal of Climate 18, 2812-2825.
http://www.cgd.ucar.edu/cas/Trenberth/trenberth.papers/massEteleconnJC.pdf
Tsonis, A.A.; Swanson, K.; & Kravtsov, S. (2007). A new dynamical mechanism for major climate shifts. Geophysical Research Letters 34, L13705.
http://www.nosams.whoi.edu/PDFs/papers/tsonis-grl_newtheoryforclimateshifts.pdf
Wyatt, M.G.; Kravtsov, S.; & Tsonis, A.A. (2011). Atlantic Multidecadal Oscillation and Northern Hemisphere’s climate variability. Climate Dynamics. doi: 10.1007/s00382-011-1071-8.
Since (to my knowledge) there’s not yet a free version, see the conference poster and the guest post at Dr. R.A. Pielke Senior’s blog for the general idea:
a) Wyatt, M.G.; Kravtsov, S.; & Tsonis, A.A. (2011a). Poster: Atlantic Multidecadal Oscillation and Northern Hemisphere’s climate variability.
https://pantherfile.uwm.edu/kravtsov/www/downloads/WKT_poster.pdf
b) Wyatt, M.G.; Kravtsov, S.; & Tsonis, A.A. (2011b). Blog: Atlantic Multidecadal Oscillation and Northern Hemisphere’s climate variability.
Important Note: While Wyatt, Kravtsov, & Tsonis (2011) are likely to stimulate a lot more discussion once a free version of their paper becomes available, it needs to be pointed out assertively & clearly that the cross-correlation approach, while informative, is patently insufficient for determining the full nature of terrestrial spatiotemporal phase relations.
Appendices
In the appendices that follow, attention is concisely drawn to key items that are consistently underappreciated in climate discussions.
Appendix A: Spatial Influence on Phase – Important
Nonrandom phase relations demand careful focus on the spatial dimension. Temporal evolution isn’t the only thing driving apparent phase.
If features grow, shrink, rotate, change shape, reflect, or move relative to the stationary windows in which they are measured, phase is affected.
The effect on summaries is plain & simple. (Anyone previously puzzled by “integration across spatiotemporal harmonics” might now get the general idea.)
Appendix B: Reversals in Temperature-Precipitation Relations
Blink between winter & summer panels of Figure 6:
Trenberth, K.E. (2011). Changes in precipitation with climate change. Climate Research 47, 123-138. doi: 10.3354/cr00953.
http://www.int-res.com/articles/cr_oa/c047p123.pdf
Temperature-precipitation relations are a function of absolutes, not anomalies. This is fundamentally important.
Insight from my local (ABC) example:
Appendix C: Global Distribution of Continental-Maritime Contrast
High-amplitude regional variance leverages global summaries, but multidecadal variations often draw misguidedly narrowed focus to the North Atlantic Ocean when it is the global distribution of continental-maritime contrast (in relation to flow patterns) that should be attracting the attention. Noting the position of the relatively small North Atlantic in this broader context, carefully compare:
1. http://icecap.us/images/uploads/AMOTEMPS.jpg
2. Figure 10 here:
Carvalho, L.M.V.; Tsonis, A.A.; Jones, C.; Rocha, H.R.; & Polito, P.S. (2007). Anti-persistence in the global temperature anomaly field. Nonlinear Processes in Geophysics 14, 723-733.
http://www.icess.ucsb.edu/gem/papers/npg-14-723-2007.pdf
Data & Symbols
‘ indicates rate of change
[ ] indicates time-integration
AAM = Atmospheric Angular Momentum
AAO = AntArctic Oscillation
ABC = Agassiz, British Columbia (west coast of Canada near USA border)
AO = Arctic Oscillation
COWL = Cold Ocean, Warm Land index
D-T = -SOI = – Southern Oscillation Index = pressure difference between Darwin & Tahiti (an indicator of El Nino / La Nina cycling)
ENSO = El Nino / Southern Oscillation
i = interannual
LOD = Length Of Day
NAO = North Atlantic Oscillation
NPI = North Pacific Index
PDO = Pacific Decadal Oscillation
PNA = Pacific North America index
PPT = PreciPiTation
QBO = QuasiBiennial Oscillation
SAM = Southern Annular Mode
SOI = Southern Oscillation Index
T = Temperature (°C)
V = solar wind speed
x = extreme
Data links available upon request.
Acknowledgements
Sincere thanks to Anthony Watts, the WUWT Moderation Team, readers, and all those who make valuable contributions towards a deeper understanding of nature.
==========================================================
PDF Version of this essay available here Vaughan, P.L. (2011). Interannual Terrestrial Oscillations (375KB)
Lots of graphs in one post is good; lots of colored lines in one graph is not good.
Probably would benefit from some animated blink-o-graphs.
That all the terrestrial relations are correlated in complicated ways simply show that they are not independent, which we would not expect in the first place. Heaping more of these on top of what is already there will make no difference at all because the inter-dependencies are still there.
This: “Multiscale complex correlation (for example using adjacent derivative based complex empirical wavelet embeddings) can measure complex nonstationary relations where simple linear correlation fails catastrophically” is just spatiotemporal mumbo-jumbo with no physical content. I shall not shy away from calling the emperor’s new clothes what they are.
The only part that fall outside that web of interdependencies is the claim: “Terrestrial phase relations with interannual [not to be confused with decadal] rates of change of solar variables, including solar wind speed (iV’), are nonrandom”. Both solar activity and atmospheric variability consist of ‘episodes’ of 1-3 years duration [with no spatial relationships whatsoever]. So any ‘multiscale complex correlations’ will always find some coherence over the limited time span we have data for because the underlying data are themselves not random. This does not mean that the data is related in any way.
I don’t pretend to understand most of the science behind this (I’m just a Mechanical Engineer), but what seems obvious to me is that all the factors involved in climate must be considered and it’s not only more complicated than we know, it’s more complicated than we can know (apologies). But I venture, with as much common sense as I can muster, the local star is the single most important variable in the whole system. Everything else must flow from that. I suppose climate could be modeled (simply and inaccurately) as a “spring-mass-damper” system with the forcing mechanism being total solar energy. Does that make some sort of sense?
I agree.
Here’s a beautiful illustration of a bunch of cycles synchronized in a non-obvious manner:
link
Obviously, CO2 aligns these variables to increase global temperature. /sarc off
If this article is a joke, it’s a lot of effort for negligible result. If it’s serious, it’s simply incomprehensible; the level of gobbledygook-jargon is intolerable. If there is something to say here, please say it in English!
I don’t even pretend to understand what the article is really getting at. It seems like a lot of “A tracks with B — except for when it doesn’t. And a lot of things all seem to change at about the same time.” Not sure what any of that means, honestly.
What language is this written in? Overbearingly Smug?
I received this from a student it would have been sent back for rewrite. Structure is wrong (fails to start small and build to a main point in a logical progression), it is aimed at the wrong target audience and assumes way to much base understanding.
Looked more like “Baffle them with BS” than a “Blind them with brilliance”
Suffers from “Look at all the pretty colors” syndrome.
And I am not sure yet what the hell the point was.
Concluding Speculation
The whole point is:
Terrestrial geostrophic balance is affected by the concert of changes in:
a) interannual (not to be confused with decadal) solar variations.
b) decadal amplitude of semi-annual Earth rotation variations – [see Vaughan (2011) & links therein].
c) solar cycle length – [see links in Vaughan (2011)].
Is that something new? Or a confirmation that their is no CO2 affect?
No offense, you’ve put a lot of work into this.
But… what are your conclusions? There’s lots of variability on various scales, does that mean anything regarding climate change? Do any of these fairly short term periodic and aperiodic fluctuations add up to anything meaningful, or indicating a long term trend?
commieBob
Great video! I think I need a drink, now, though…
One of the reasons I wanted to post this was to see the reaction. Whatever you’re thinking, please say it.
I’m particularly curious to hear from the Tsonis, Swanson, & Kravtsov (2007) fans […and we know there are a good number of them around].
Thanks sincerely to those who have taken the time to comment. I look forward to carefully reading what others have to say.
I also have a request of the community. There’s a slide in Jasper Kirkby’s recent SFU IRMACS presentation that I need to find in the literature. (I’ll look up the time-index tomorrow…)
Best Regards.
I’m sorry, but you lost me with “Elegant factor analyses by Trenberth…”
Putting the word “elegant” in the same sentence as “Trenberth” seems contradictory, unless in the context about “Trenbreth’s elegant emails written to cover up climate skullduggery” or some such.
http://wattsupwiththat.com/2011/01/13/trenberths-upcoming-ams-meeting-talk-climategate-thoughts/
Paul,
May I be so bold to say that you are on to something…but, be that as it may, the Newspeak interannual blah blah blah stuff…has got to go.
In science, as you well know….might does not make right,
As a matter of fact…it often creates error.
Just by nomenclature alone, you open yourself up to criticism from the wolves….such as Dr. S.
And actually, besides rigorous tests…their criticism is is well-founded. Nobody knows what the hell you are talking about. You have to define it more clearly.
Your friend,
Chris
Norfolk, VA, USA
Leif Svalgaard wrote (May 15, 2011 at 5:41 pm):
“That all the terrestrial relations are correlated in complicated ways simply show that they are not independent, which we would not expect in the first place. Heaping more of these on top of what is already there will make no difference at all because the inter-dependencies are still there.”
Well put Leif.
Now try to explain that to climate science & statistical science, which insist on drawing so-called “conclusions” based on PATENTLY FALSE assumptions of independence.
You nicely make my PRIMARY point.
Leif Svalgaard wrote (May 15, 2011 at 5:41 pm):
“This: “Multiscale complex correlation (for example using adjacent derivative based complex empirical wavelet embeddings) can measure complex nonstationary relations where simple linear correlation fails catastrophically” is just spatiotemporal mumbo-jumbo with no physical content. I shall not shy away from calling the emperor’s new clothes what they are.”
Can you be more specific about what you find so offensive about Maraun & Kurths (2005) section 3 and related methodologies?
I’m not a scientist either, but I think I got a conclusion from this (accepting that it could have been written more clearly): standard linear analysis fails to recognise correlations between between non-random cyclical events because it cannot take account of the fact that, in a spatio-temporal chaotic system, the changes you are measuring move around. Sometimes therefore, your data will appear to be out of phase when it is in fact in-phase.
Perhaps this is something all other readers of WUWT were aware of, but it wasn’t something I’d thought of before!
CRS, Dr.P.H. says: “Putting the word “elegant” in the same sentence as “Trenberth” seems contradictory…”
Well, I beg to differ. Here are two well-known Trenberthisms:
“I remain somewhat skeptical… It is hard to make data where none exist.” (regarding Steig et al)
“The fact is that we can’t account for the lack of warming at the moment and it is a travesty that we can’t.” (from the Climategate emails)
These quotes are brief, pithy, very much to the point, and accurate. In short, elegant.
“absolutes not anomalies”
i love it!
Paul,
Does the variation represented in the vertical axis represent the state of the correlation between the indices. If not, then what?
The link to Pielke Senior between the first and second charts did not work for me.
~~~~~~~~~~~~~~~~~~
Next, I think this post fits nicely with Donald Rumsfeld’s explanation of knowns and unknowns. Namely . . .
“Lack of widespread awareness of the . . .
I am aware of the Pacific Decadal Oscillation (PDO) but I do not understand why it exists, what it is doing, nor why it changes when it does. The same can be said for many of the other oscillations mentioned, and thus this is somewhat of a “paralyzing bottleneck” toward improving my personal know, don’t know, don’t know that I don’t know list of things. The post does inform me that there seem to be a lot of known (by somebody) things that I do not know about “the spatiotemporal nature of interannual terrestrial oscillations” and will never know much about most of them.
That seems to be the general purpose of the exercise and in that sense it has succeeded.
I read it backwards on tape and Bill Clinton said “I did not have sex with that woman”.
Paul Vaughan says:
May 15, 2011 at 10:16 pm
Now try to explain that to climate science & statistical science, which insist on drawing so-called “conclusions” based on PATENTLY FALSE assumptions of independence.
I don’t think they are doing that. It seems to me that they are well aware of the fact that almost all measures of atmospheric variability are strongly correlated with each other and hence not independent.
Paul Vaughan says:
May 15, 2011 at 10:27 pm
Can you be more specific about what you find so offensive about Maraun & Kurths (2005) section 3 and related methodologies?
I don’t find them ‘offensive’. I fail to see [could simply be a failure on my part] what they bring to the table that is not trivial, new, and interesting. What do we learn that we did not already know? Where is the phenomenon that ONLY can be discovered with their methodologies? and that will not yield to standard techniques.
Now, in my field, we deal with the ultimate in spatio-temporal analysis: successful helioseismic exploitation of the integrated effect of millions of standing waves observed simultaneously over the entire solar disk. We don’t have any difficulties with what you call ‘abstract conceptualization’ [although it is not quite clear to me what you mean]. I don’t think people that search for oil with seismic techniques have difficulties either, so I fail to see how the formalisms you are pushing will improve something that works quite well already.
Paul
Your article may be a goldmine for more enthusiastic reader, but I felt overwhelmed by methodology of the presentation.
Inter annual oscillations very often have appearance of random noise. One could consider this to be a case with the Multivariant ENSO or the closely related Southern Oscillation Index (SOI).
My short excursion into analysis of these events has found a close correlation with a non-climate process with power and the means to influence, or even drive ENSO.
http://www.vukcevic.talktalk.net/SOI.htm
Paul: I, too, found the post somewhat impenetrable. I decided not to comment after seeing some of the responses, lest I seem to be piling on. But since you ask:
(1) I’m not a fan of “wiggle matching,” especially of the sort where “it matches except where it doesn’t.” There must be some sort of statistical or mathematical handling of in-phase / 180° out-of-phase periodic data to make such statements meaningful.
(2) I really like graphs and charts that have a full legend: Ordinate and abscissa units, indication of any averaging/smoothing and how it was done, base period for anomalies, labels, all of it, the whole nine yards.
(3) I have a strong preference for authors who eschew sesquipedalian phraseology. I’m not accusing you of being inebriated by the exuberance of your own verbosity, but rather of having missed an opportunity to achieve even greater clarity.
(4) I’d suggest adopting the methods of the award-winning teacher who explained her pupils’ outstanding performance by saying: “First, I tell ’em what I’m gonna tell ’em; an’ then I tell ’em; and then I tell ’em what I told ’em.”
(5) I read the charts carefully and thus was looking for the indicator of derivatives, which I found only because I knew it must be there. Others might have easily missed the little ‘ symbol. It would be easier to read (though more challenging to include) the full dx/dt nomenclature. Legibility helps.
Hope this is helpful. I’m a technical editor, so this sort of paper isn’t as strange to me as it is to some, but I still found it too challenging to translate in the time I had available this weekend.
Jorge Kafkazar, MS
Rather beyond my comprehension level. I notice there seems to be at least four time periods. I assume the vertical axis is common to all graphs. No individual trace appears to exceed the +1 -1 levels over the total 110 year period. I have to ask, “What or where is the problem?”
Paul Vaughan says:
May 15, 2011 at 10:16 pm
Leif Svalgaard wrote (May 15, 2011 at 5:41 pm):
“That all the terrestrial relations are correlated in complicated ways simply show that they are not independent, which we would not expect in the first place. Heaping more of these on top of what is already there will make no difference at all because the inter-dependencies are still there.”
Well put Leif.
Now try to explain that to climate science & statistical science, which insist on drawing so-called “conclusions” based on PATENTLY FALSE assumptions of independence.
You nicely make my PRIMARY point.
==========================================
lol, well why didn’t you say so in the first place!
Seriously though, Paul I thank you for all of this information. Its going to take me a while to sift through much of this. I can appreciate all the effort you went through to make these statements. While, like others here, the presentation style was a bit much for me, I acknowledge that we all respond to visuals in a different manner. But what seems lost on many here, is that the concluding statement is often the most poignant.
We have a lot of work to do (both exploratory & methodological).
Yes, and it is seemingly much more complex than any one can grasp.
Thanks much,
James
At night I can hear my watch ticking, very close to my ear on the side table. I can also hear my wife’s clock ticking on the other side of the bed.
Lots of the time, they tick together, making a nice rhythm. Then after a while, the patterns breaks down, with apparently random gaps between ticks. Then shortly after, they seem to re-sync and tick along together for a while.
So how do you explain this strange phase shift, when they appear to by in sync for substantial periods, then the odd random periods before they re-sync?
Or perhaps they are simply both ticking happily along in their own way, and I’m imposing a correlation where none exists?
Addition to Data and Symbols
AMP = Above My Paygrade
I need to go and lie down.
Maybe this is what Leif was getting at, but I’m not sure….
I have to wonder what these plots would look like for a variety of time series of low-pass filtered random numbers. Given only a certain number of climate events/quasi-cycles in various regions in the last 100 years, there can be apparent associations that have no physical connection. For example, you can do spectral analysis of time series of random numbers and get spectral peaks that have no physical significance.
Multi-variable perambulations of non-interconnected newspeak in a close formation of unintelligible graphic representations of a somewhat dubious usefulness with no apparent conclusion. This post may excite perhaps a dozen people in the entire world , until it is rendered into the laymans dialect it will be an abstract of non-sense.
Paul, I was about to post but Jorge Kafkazar sums up my main complaints about this presentation:
(2) I really like graphs and charts that have a full legend: Ordinate and abscissa units, indication of any averaging/smoothing and how it was done, base period for anomalies, labels, all of it, the whole nine yards.
This is a scientific subject, posting sociologist style graphs is really of little help. I’m very surprised that someone with a M.Sc would present anything in that way. I can’t see what most (any) of these graphs are supposed to be showing and I don’t have the time play guessing games. If you have something interesting, please state exactly what your point is and correctly and fully label your graphs.
“data available on request”.
Why not just post a link with each graph? You spent the time plotting and writing all this , what is so hard about providing a link?
Also any discussion is near impossible if you don’t label each graph , figure 1. etc.
Almost any semi random data starts to look like it has cyclic components once you smooth it. Any climate data will have interannual variations. If the go in and out of phase it means they have different frequencies. There is nothing interesting in that.
I can only guess that you have failed to get across what you were hoping to show.
best regards.
A thesis superviser 50 years ago told me that the best way to hammer in a nail is to hold it, give it a few taps, then drive it in with 2 or three strong blows, after I had given him everything, including the kitchen sink in support of an idea. The enthusiasts who spend any time at WUWT include climate scientists, other scientists and engineers, but also, the whole gambit of laypersons – I’ve read pithy comments from farmers, politicians, military pilots, journalists … and so on. They all seem to find the subjects here interesting so it is a generally bright lot. If you have a legitimate thesis, it can be put forward intelligibly to all. You can of course put in a few highly technical asides and links for the expert, but even the expert wants clear exposition.
Leif Svalgaard wrote (May 15, 2011 at 11:39 pm):
“It seems to me that they are well aware of the fact that almost all measures of atmospheric variability are strongly correlated with each other and hence not independent.”
So why the culturally dominant adamant insistence on statistical inference based on the FALSE assumption of independence?
Anthony said “timing is everything”.
Actually:
Timing & spacing.
Timing AND spacing.
This one comment makes the whole exercise worthwhile:
John Syfret wrote (May 15, 2011 at 11:00 pm):
“I’m not a scientist either, but I think I got a conclusion from this (accepting that it could have been written more clearly): standard linear analysis fails to recognise correlations between between non-random cyclical events because it cannot take account of the fact that, in a spatio-temporal chaotic system, the changes you are measuring move around. Sometimes therefore, your data will appear to be out of phase when it is in fact in-phase.
Perhaps this is something all other readers of WUWT were aware of, but it wasn’t something I’d thought of before!”
Appreciated – Thanks John.
Roy Spencer wrote (May 16, 2011 at 4:57 am):
“Maybe this is what Leif was getting at, but I’m not sure….
I have to wonder what these plots would look like for a variety of time series of low-pass filtered random numbers. Given only a certain number of climate events/quasi-cycles in various regions in the last 100 years, there can be apparent associations that have no physical connection. For example, you can do spectral analysis of time series of random numbers and get spectral peaks that have no physical significance.”
Unwindowed spectral analysis is of limited utility in part for the reason’s Roy explains, but I assure Roy that randomly generated numbers (even smoothed ones, contrary to popular misconception) do NOT routinely fool methods such as those suggested in Maraun & Kurth’s (2005) section 3 …nor do they routinely produce plots like these
http://wattsupwiththat.files.wordpress.com/2011/05/vaughn_npp_image6.png
http://wattsupwiththat.files.wordpress.com/2011/05/vaughn_npp_image5.png .
Leif is saying the variables are NOT independent because they ARE physically connected. My question is: Why do climate scientists pretend the variables are independent when conducting statistical inference?
For readers lacking background but trying to follow along:
Fundamentally unlike data exploration, statistical inference is based on ASSUMPTIONS — assumptions which are far too often made unconsciously. Worse than that, the establishment INSISTS upon them, even when they are patently untenable. This is unacceptable.
oldgamer56 wrote (May 15, 2011 at 8:18 pm):
“it is aimed at the wrong target audience”
Astute observation.
Addressing the “wrong” audience was sure to draw misunderstanding & vicious collateral attack. Isn’t it better to let climate scientists know that some of us know that their uncertainty estimates are based on patently untenable assumptions? Isn’t that worth risking having one single neck turned into a red sprinkler by the “wolves” (as Chris calls them)?
Paul Vaughan says:
May 16, 2011 at 7:20 am
Leif is saying the variables are NOT independent because they ARE physically connected. My question is: Why do climate scientists pretend the variables are independent when conducting statistical inference?
I don’t think they pretend that. Do you have some papers or statements by climate scientists where they say that they pretend the variables are independent? In most cases that I know of, they are trying to show that there is some dependency, e.g. that temperature depends on CO2, or that rainfall in some region depends on ENSO or some such.
In Jasper Kirkby’s SFU IRMACS video from time-index 35:45 to 35:50 he’s drawing attention to a Southern Ocean wind/NaCl pattern (right panel).
Can anyone point to the journal article?
This appears to be the “missing link” between Appendix C (above) and Trenberth, Stepaniak, & Smith’s (2005) powerful Table 3 & Figure 3.
The role of north-south terrestrial asymmetry has been underappreciated. Some of you may recall this:
http://wattsupwiththat.files.wordpress.com/2010/08/vaughn_lod_amo_sc.png
http://wattsupwiththat.files.wordpress.com/2010/09/scl_northpacificsst.png
http://wattsupwiththat.files.wordpress.com/2010/09/scl_0-90n.png
SCL’ = rate of change of solar cycle length = solar cycle deceleration
This detour into “interannual terrestrial oscillations” is just a side trip to help those blinded by interannual spatiotemporal chaos to patterns illustrated by Wyatt, Kravtsov, & Tsonis (2011) https://pantherfile.uwm.edu/kravtsov/www/downloads/WKT_poster.pdf .
John F. Hultquist wrote (May 15, 2011 at 11:35 pm):
“The link to Pielke Senior between the first and second charts did not work for me.”
Some issue with the apostrophe ‘ in the title. Try this approach:
http://pielkeclimatesci.wordpress.com/?s=Atlantic+Multidecadal+Oscillation+and+Northern+Hemisphere+Wyatt
Interesting that the article hasn’t attracted more attention from WUWT commenters, given how excited many are about “60 year cycles”.
“”””” Elegant factor analyses by Trenberth, Stepaniak, & Smith (2005) concisely chart the limits of linear climate exploration, providing strong clues that the North Pacific is a globally pivotal intersection. “””””
I nominate the above on behalf of its self referencing authors, for entry into this year’s Bulwer-Lytton prize competition; aka the Bloviating BS prize.
“”””” commieBob says:
May 15, 2011 at 6:27 pm
Here’s a beautiful illustration of a bunch of cycles synchronized in a non-obvious manner:
link “””””
Well commie, I watched the video several times, and I saw no evidence of any synchronisation. As imple demonstartion that the period of a simple pendulum varties with its length, and a camera gimmic to show it off.
Now it would have been far more illuminating, if they had included some sort of energy coupling between the several pendula to show how they exchange energy among themselves. Coupled penula are much more interesting than length varied ones.
erlhapp wrote (May 15, 2011 at 11:35 pm):
Paul,
Does the variation represented in the vertical axis represent the state of the correlation between the indices. If not, then what?”
No. The interannual filter used simply contrasts years with immediately-adjacent years for each variable. Amplitude is normalized to facilitate visualization.
Erl, I’m wondering if you are seeing some connection between this [ http://wattsupwiththat.files.wordpress.com/2011/05/vaughn_npp_image8.png ] and your work on polar voritices & geomagnetic Dst index?
Btw, there’s a whole decade where local precipitation in my area perfectly tracks geomagnetic indices. I want to know why and I’m not buying this nonsense that it’s just random chance.
Leif has to prove that A, B, & C swirl iV’ while X, Y, & Z (independent from A, B, C, & iV’) swirl iNPI’ to satisfy me that he is correct in ruling out a link between iV’ and tropospheric variability. I will welcome a convincing argument, but I’m skeptical that one can be produced.
Thanks for joining the discussion Erl.
Leif Svalgaard wrote (May 16, 2011 at 7:43 am):
“I don’t think they pretend that. Do you have some papers or statements by climate scientists where they say that they pretend the variables are independent?”
They’re a dime a dozen. It’s standard widespread practice – even forced upon subordinates – to assume independence. It’s often done unconsciously in statistical inference; most don’t diligently retain cognizance of the model assumptions (for example when interpreting).
I’m not speaking just of climate science. This is a widespread systemic problem in our society that produces LINEAR p-values to nonsensically dismiss correlations of the following (complex) variety as “insignificant”: http://wattsupwiththat.files.wordpress.com/2011/05/vaughn_npp_image3.png
This is fundamentally unacceptable practice.
“”””” Roy Spencer says:
May 16, 2011 at 4:57 am
Maybe this is what Leif was getting at, but I’m not sure….
I have to wonder what these plots would look like for a variety of time series of low-pass filtered random numbers. Given only a certain number of climate events/quasi-cycles in various regions in the last 100 years, there can be apparent associations that have no physical connection. For example, you can do spectral analysis of time series of random numbers and get spectral peaks that have no physical significance. “””””
Right on Roy; apply random white noise to any filter input, and it will output that part of the sepctrum which elicits a response, leading to the eroneaous conclusion that it is a signal.
Which begs the question.
In talking about correlation which Dr Svalgaard alluded to; do climatists distinguish or differentiate between the simple correlation of the numbers (mathematics) that represent the observations of climate phenomena; and the physical correlation of the pehenomena themselves.
To me, the statement that experimental observation (A) and experimental observation (B) are “correlated”, implies (or if you prefer; I “infer” from that), that there is some underlying cause and effect relationship between (A) and (B) or alternatively between (A and B) and some other (C) (which might be the cause of both (A and B) ).
That the numbers used to represent each phenomenon (A,B, C….) show some non-zero mathematical cross-correlation coefficient, is not of itself evidence that the phenomena are related let alone correlated. Absent a cause and effect connection; we merely have demonstration of the non-zero bandwidth of the analysing filter; that lets several unrelated signals through.
Paul Vaughan says:
May 16, 2011 at 9:00 am
Leif has to prove that A, B, & C swirl iV’ while X, Y, & Z (independent from A, B, C, & iV’) swirl iNPI’ to satisfy me that he is correct in ruling out a link between iV’ and tropospheric variability. I will welcome a convincing argument, but I’m skeptical that one can be produced.
The shoe is on the other foot. The one claiming a link has to produce the proof. The null-hypothesis is no link. The main arguments [which are not by themselves ‘proofs’] against a link are the weakness of correlations, the lack of mechanism, and insufficient energy input.
Paul Vaughan says:
May 16, 2011 at 9:11 am
“I don’t think they pretend that. Do you have some papers or statements by climate scientists where they say that they pretend the variables are independent?”
They’re a dime a dozen.
I’ll be content if you show me just one. As I said, most people are assuming that the variables are dependent, otherwise why investigate relationships between variables that do not depend on each other. And in the natural sciences, statistics never proves anything, just serves as an indication of a possible line of investigation into the physics and mechanisms involved.
Two points:
The first is that many statistical signal processing techniques require an ensemble approach,, i.e.: the behaviour of the means of some parameter. We only have one climate record and therefore we cannot use an esemble approach without segmentation of the data and thus limit the minimum observable frequency. I agree with Spencer’s point about the specrum of a low pass filtered random variable. This is easy to show as random provided one has enough data. With a single observation, statistical methods to eliminate the effect of apparant non-randomness is not robust.
The second is that any interaction between systems of different frequencies will inevitably be non-linear, common examples are entrainment in which a non-linear oscillator is ” captured” by an entraining signal or modulation. I would agree that the assumptions of statistical independence and linearity coherence/phase are probably incorrect in the climate as, I presume, the various measures discussed are physically connected. If this is done on a routine basis, I suggest that some remedial mathematics for the perpetrators.
I may have missed the point but, as a signal processer, I found this paper fairly opaque. I feel an expanded version of the methods and axis labelling would lead to greater “coherence” in the presentation.
You don’t have to know how the electrons are flowing through the chips to know what a computer can do!
From my years of electronics communications work modulating and then demodulating signals that have been transmitted over a radio frequency carrier method, I can see ways the overall long term signals in the weather and climate data could be understood.
If the longest wave pattern indicative of a composite wave pattern is used as the key for the demodulation of the packets of individual periods of oscillation patterns, then the over lapping of several periods of these patterns should give an indication of the reproducibility of the original signals.
The inner planets and the outer planets have their own separate periods of harmonic oscillation as a result of billions of years of gravitational, tidal and electromagnetic interactions that are in a resonate pattern, the inner planets have a beat frequency of about 6558 day period with the outer planets settled into a period of somewhere between 172 and 179 years.
Most of the cyclic patterns seen in the above graphs are a result of the shorter term periods of interactions of both the inner planets and the outer planets beating together and generating these short term affects in the global circulation patterns.
As RC Saumarez says:
May 16, 2011 at 10:14 am
“We only have one climate record and therefore we cannot use an ensemble approach without segmentation of the data and thus limit the minimum observable frequency.”
To solve this problem I looked at three patterns of the 6558 day period, overlaid them at the daily weather data level, and plotted the resultant combined signal for Precipitation, and temperature patterns for the USA, extended that cyclic interpenetration for a six year period, and plotted out maps to show the repeating reoccurring patterns in the global circulation, as a ( 6 year long stretch, we are now ~40 months into the posted 6 years long) forecast for part of the current repeat of the 6558 day long cycle.
The precipitation patterns repeat well enough to show the locations of the repeat of the 1974 tornado out break we had this spring, and will also show the repeat of the severe flooding in Australia. The problem with this forecast technique is the outer planet ~172/179 year pattern is out of phase with the inner planet harmonics so it needs to be compensated for as well.
The best method I have come up with so far would be to plot the data excursions from the expected past 6558 day patterns, by looking at the 5 to 10 day long disruptions caused by the current 6558 day period heliocentric (or Synod) conjunctions, as the corrective algorithms to adjust the inner planet patterns for the effects of the outer planet modulations.
The Earth has a synod conjunction with each of the four major gas planets with a little over a year between impulses, (~+2 days for Neptune, ~+4 days for Uranus, ~+13 days for Saturn, ~+34 days for Jupiter) these past almost annual periods could be segmented and applied into the composite forecast as an adjustment algorithm to further enhance the pattern match, to increase the accuracy (~85%) of the inner planet 6558 day pattern to closer to 95% with the addition of the compensation for the outer planet interactive influences.
I am in the process of rewriting the map making software to fine tune the resultant process to get a better repeatability of the temperatures as well. With the next posting of updated several year long series of forecast maps I will be including the areas of Alaska, Canada, and Australia (with enough hind cast [+6 months] to show how well it did with the flooding there). I will be adding a fourth cycle going back another 6558 day period to use the additional analog year starting in ~1938 for 2011.
If any of the readers here have helpful suggestions on how this could be better accomplished, I would like to hear from them.
Paul
My brain has tried to work out what your article is about and what it has come up with is that there are many things affecting the climate and to try and remedy one of them is a waste of time.
PS glad you can spell Vaughan correctly.
Paul says:
“The interannual filter used simply contrasts years with immediately-adjacent years for each variable. Amplitude is normalized to facilitate visualization.”
Now it starts to make sense even though I have not got the first clue as to how the data is generated.
My take:
Mathematical techniques to check associations can get us only so far. Even if the association is perfect we have to know why, i.e. what the driver is. I can help in that respect and answer the question that Paul poses in this way: Erl, I’m wondering if you are seeing some connection between this [ http://wattsupwiththat.files.wordpress.com/2011/05/vaughn_npp_image8.png ] and your work on polar voritices & geomagnetic Dst index?”
Yes, I can. The polar circulations couple the troposphere and the stratosphere. In the southern hemisphere the result is a zone of very low surface atmsopheric pressure at latitude 60-70 degrees south due to the descent of ozone into the troposphere, the resultant warming and reduced weight of the atmospheric column. The ‘annular ring’ of lower surface atmsopheric pressure is amost complete (ring like) but it is broken by the influence of the landmasses of the southern hemisphere. In the southern hemsiphere the flux in ozone into the tropsophere is continuous throughout the year but the background levels of ozone are comparatively much lighter than in the northern hemsiphere and the influence on cloud cover (as inferred from change in sea surface temperature) is perceptible only at higher latitudes.
In the northern hemsiphere ozone descent into the troposphere occurs in the main in northern winter because it is at this time that the troposphere and stratosphere become a strongly coupled circulation. The coupling depends upon the estabishment of a regime of air temperature declining continuously with increasing elevation into the mid-stratosphere, favouring convection, as it does in the troposphere.
Because the Arctic stratosphere has relatively high levels of ozone (relatively unaffected by coupling in summer) the effect on cloud cover (as inferred via change in sea surface temperature when compared with flux in surface pressure in the main zones of ozone descent) is much more dramatic, more obviously synchronised and coherent than it is in the southern hemisphere and the influence extends as far as the mid latitudes of the southern hemisphere.
Proving all this in a satisfying way is not something that can be done in this post but I refer readers to http://www.happs.com.au/images/stories/PDFarticles/TheCommonSenseOfClimateChange.pdf
In that work I also deal with the connection between geomagnetic indices and surface pressure and the way in which the polar votices affect backround levels of ozone in the polar stratosphere.
Now, I want to get the the main thrust of Pauls post that suggests that the North Pacific is pivotal in the process of short and long term climate change.
The pattern of descent of ozone into the troposphere in the northern hemsiphere is affected by the distribution of land and sea in just the same way as in the southern hemisphere. But there is more land and less sea. The upshot is one zone of descent in the north Pacific and another across the north Atlantic.
What I am referring to here is what has been described by Baldwin, Dunkerton and others only since the turn of the century as the northern and the southern annular modes, recognized as associated with interannual climate variability. But so far ‘we know not how’. In other words, the maths points to the association, the spato-temporal aspects are well described but the understanding of process is as yet so slight as to be insignificant.
Climate change is driven from the poles. The most obvious links are via the NAM (Northern Annular Mode), that is synchronous with the Arctic Oscillation. It has long been recognized that there is a close relationship between solar indices and the Arctic Oscillation in northern winter months. In my work I reference Palamara in that respect.
The weakness of all correlative work comes down to lack of understanding of process. I suggest that this situation will change as we build on the current understanding of the NAM and the SAM and understand the importance of convection in the polar atmsophere and it’s importance in driving stratospheric ozone levels and tropospheric cloud cover.
CommieBob’s video link is highly explanatory.
The simple pendulum experiment shows that oscillators of any type can show all sorts of correlations that are purely artifacts of the point of view. Look at the pendulum experiment from the side and you simply see 16 pendulums swing side by side at different frequencies. If they had panned around the platform at other angles other apparent correlations or patterns would have appeared.
It’s related to Steveta_uk’s tick-tock cycle generated by two watches. Or the beautfiful Moire’ patterns you can see through layers of silk. The Moire’ patterns come from minute guage and thread diameter variations in the silk and the angle of the lighting. Neither one contains any information except about the frequencies involved. Theoretically, you could take the period of the tick-tocks and calculate the exact relative frequency of each watch. You could model the spacial Moire’ patterns and estimate the guage and diameter variations of the sllk. But the patterns themselves don’t carry any information about the cause of the oscillation or pattrern
This paper has the same problem. Any kind of correlations between oscillators don’t tell us anything about the mechanism of what is actually going on. In order to get that you would need an accurate model of say the Pacific Decadel Oscillation- where the ocean flows, temperatures, winds, etc., are known and then extend it to include how the various ocean basins interact. Not to take shots at Leif Svalgaard, but the correlations don’t really even show that the items are interdependent. To do that you would have to show how they are connected, which is more than mere correlation.
George E. Smith says:
I saw the video first on http://www.boingboing.net . Here’s a quote.
In other words, each pendulum has a subharmonic whose period is one minute.
commieBob says:
May 16, 2011 at 6:06 pm
From the shortest pendulum to the longest, each pendulum is one swing per minute less than the one before it. The effect is a hypnotic dance that repeats every 60 seconds.
But which does not demonstrate synchronization at all, as there is no coupling between the pendulums. The dance is totally artificial and contrived.
Richard Holle
I look forward to reading your work, particularly releating to Australia.
Paul Vaughan
I realise that the following will be hurtful, BUT:
I suggest that you give your paper to someone who is expert in communication.
It needs to be completely re-written, showing exactly what you are criticising (not just saying that there are many examples) and explaining how and why your method is superior.
But please, do not attempt this task yourself.
Leif, I [not you] decide what will convince me that iNPI’ & other i-waves are independent of iV’. As indicated above, I’m willing to be convinced, but I’m skeptical that anyone can produce a convincing argument.
As for the other “misunderstanding”:
You know very well the following:
“Let X1, X2, X3, … ~i.i.d.”
You also know very well the demands for some kind of formal statistical inference in formal pubs.
Linear methods alone aren’t sufficient for inference on these variables. There are plenty of methods that COULD be developed that might lead to far more meaningful p-values and confidence intervals. One sees the pioneering going on, but it’s not all the way there yet.
A merger of something like Maraun & Kurths hybridized with cross-wavelet methods and VEOF is one suggestion I’ve pitched to a local expert on nonlinear factor analysis. Unfortunately, he’s retiring. Currently I lack the resources to pursue the project independently, but I’ve developed a prototype.
After developing the prototype, I decided the audience here would not understand the algorithm output — hence the decision to post simple graphs and see if folks’ intuition is enough to see what the algorithm sees. The comments provide valuable feedback. I find it particularly interesting that most of those sufficiently motivated to comment appear more concerned with formal cosmetics than informal climate discussion.
Erl, thanks for your interesting comments.
Thanks to all who have commented.
I fully expected serious misunderstandings & vicious feedback on this informal article.
I firmly retain my opening assertion:
“Lack of widespread awareness of the spatiotemporal nature of interannual terrestrial oscillations is perhaps the most paralyzing bottleneck in the climate discussion.”
Leif Svalgaard’s first paragraph in his first post and the comment from John Syfret assured me: Mission accomplished on the Pareto Principle with anticipated & entirely acceptable collateral damage.
Best Regards to All.
Paul Vaughan says:
May 16, 2011 at 8:25 pm
Leif, I [not you] decide what will convince me that iNPI’ & other i-waves are independent of iV’. As indicated above, I’m willing to be convinced, but I’m skeptical that anyone can produce a convincing argument.
As Richard Feynman once remarked, the easiest one to fool is oneself. You have not even described what the relationship is. Suppose that the solar wind speed depends on the day of the year. What would that do to the ‘relationship’? with the seasons being a confounding variable…
This comment is addressed to all those who have criticised this post on the basis that correlation is fortuitious, and unrelated to causation. It is also addresssed to those who find the presentation obscure and those who maintain that ‘wiggle matching’ is unlikely to bear fruit.
The task of identifying the cause of climatic variations that are sometimes described as ‘natural variations’ involves demonstrating a plausible mechanical link between variables that is consistent. What we have in this post is simply a demonstration of the consistency, or lack thereof in the links between certain indices that are known to relate to variations in temperature and rainfall in some locations on the globe.
If a pattern of variation is consistent for a period of time and becomes inconsistent it possibly represents the waxng and waning influence of second or third drivers. For example, there are two poles and the influence of these poles changes over time.
It is easy to criticise the work because of what it is not. The work makes no attempt to describe modes of causation although it is suggested that iV, the variation in the solar wind could be involved. As I intimated above, that suggestion is not novel, so far as it relates to the northern annular mode, also described as the Arctic Oscillation. But, here it is shown that variation in iV relates to more indices. That’s progress. No work could survive scrutiny if it could be criticised on the basis that it does not provide a complete picture of the world. So, much of the criticism here is irrelevant.
Matching the wiggles is the first observational task. A mathematical technique that enables one to identify the degree of symmetry is valuable.
The second task is to explain why the wiggles match. It would be nice to have the two together but it is not always possible.
Leif does not appear to contradict the proposition that the wiggles match in a way that suggests they could be related. That is progress indeed.
Others will suggest why the wiggles relate as they do and their work is made easier to the extent that the degree to which variables move together has been identified.
Then there is the sort of criticism offered by Ausie Dan, which is just posturing, egotisitical, and hurtful. No place for that at all. A really cheap shot. A stone thrower getting into the act.
I look in vain for the person who asks the simple question: Why is it so? When that question is asked the real discussion begins.
Leif, whether relationships are physical or not, they are still of interest to someone who catalogs them. Awareness of confounding is preferable to ignorance of it. I understand your concern that some innocents will misunderstand. I also acknowledge that physicists define “relationship” differently than do statisticians.
Eloquent words Erl (erlhapp May 16, 2011 at 9:43 pm).
Fyi: The relationship with iNPI’ is much tighter & much more systematic than with iAO’ & iNAO’ (or iNAM’ if you prefer).
Thanks Paul,
Info at http://www.cgd.ucar.edu/cas/jhurrell/indices.info.html#np is useful.
On solar and terrestrial links
http://www.vukcevic.talktalk.net/CD.htm
with more details to follow.
Leif Svalgaard says:
May 16, 2011 at 9:15 pm
You have not even described what the relationship is. Suppose that the solar wind speed depends on the day of the year. What would that do to the ‘relationship’? with the seasons being a confounding variable…
Paul, you ducked this question.
erlhapp says:
May 16, 2011 at 9:43 pm
The work makes no attempt to describe modes of causation although it is suggested that iV, the variation in the solar wind could be involved. […]
Matching the wiggles is the first observational task. A mathematical technique that enables one to identify the degree of symmetry is valuable.
Except that no attempt was made to quantify the degree of wiggle matches.
Leif does not appear to contradict the proposition that the wiggles match in a way that suggests they could be related. That is progress indeed.
Let me try to go behind ‘appearance’. I don’t think the wiggles match enough to be of interest. Since Paul ducks quantifying the claimed match, no progress seems possible.
Paul Vaughan says:
May 16, 2011 at 10:26 pm
I understand your concern that some innocents will misunderstand.
We are all ‘innocents’ in this game, except the guilty ones.
I also acknowledge that physicists define “relationship” differently than do statisticians.
There is a strong relationship between shoe size and reading ability among school children. A statistician concludes that bigger feet helps the child to learn to read. As I said, all the other graphs [apart from the one with iV] presented were just correlating climate with climate so are not of any relevance or interest. The iV ‘relation’ is not quantified and the solar wind speed time series does not seem to have any spatial aspect to it to merit the notion of ‘spatiotemporal’.
So, it would perhaps be of interest to examine the iV graph in detail.
1) Where does the solar wind data come from?
2) How is the data smoothed? precisely, how is the techniques of section 3 of Maraun & Kurths applied to the data?
3) Can you quantify the non-randomness?
4) Address the confounding aspect with solar wind speed depending on time of year
Leif Svalgaard falsely asserted (May 17, 2011 at 4:08 am):
“Except that no attempt was made to quantify the degree of wiggle matches.”
Correction to the statement:
Multiscale complex correlation has not been presented in this article, but it has been measured by 2 different methods.
The annual & semi-annual signals (phase-locked to the terrestrial year) are on my radar, as are the dependence of your V reconstruction on space and the skewed distribution of your proxy.
–
Leif Svalgaard also falsely asserted (May 17, 2011 at 4:08 am):
“A statistician concludes that bigger feet helps the child to learn to read.”
Here you’re just having fun playing to the audience.
Some stats profs go to a fair amount of trouble cautioning their students about how to write conclusions that are technically correct. The most common error is failure to acknowledge conditional dependence of interpretations on null hypothesis model assumptions. In your example it would be technically correct to note a relationship between shoe size & reading skill. In an introductory level course, one might take the opportunity to reinforce lessons about confounding & lurking variables.
—
I developed my own methods independently well before I ever read Maraun & Kurths (2005). When I found the paper (& its section 3 in particular), it became a convenient link. Same comments would apply to Schwing, Jiang, & Mendelssohn (2003) [except that the free version of their paper has disappeared from the web].
I explained the filtering above (in response to a question from Erl). I can add that the normalization is based on the maximum absolute deviation.
Pursuing p-values & confidence intervals would be premature at this stage. More thorough data exploration is prerequisite to sensible statistical inference (contrary to dominant mainstream pressures that ignore Simpson’s Paradox).
vukcevic, I look forward to reading the article you’re assembling.
NPI:
http://www.cgd.ucar.edu/cas/jhurrell/Data/npindex.mon.asc
Leif Svalgaard says:
May 16, 2011 at 7:48 pm
Almost every waveform I have ever looked at was totally artificial and contrived. 😉
It doesn’t matter if the pendulums are actually coupled or not. If we were to actually synchronize them with a clock pulse whose period was one minute, their oscillations would look the same (except they might not be damped).
Paul Vaughan says:
May 17, 2011 at 7:35 am
…………….
Paul
Some scientists often get into ‘gainsay’ arguments to the extent where further exchange of views is counter productive.
What I will be presenting is a natural ‘down to Earth’ likely cause for the climate change based on solid data. May be a fluke, but since it is equally applicable to the North Atlantic as well as the North and Central Pacific, hopefully there is more to it than just a wiggle matching. It may be a bit controversial, not for reasons related to the climate change, nothing odd or unknown there, all strait forward stuff, but for revealing couple of unrelated ‘known unknown’ spoilers.
There are some interesting comments from Scafetta on the other thread.
Novel methods of analysis can be expected to invite miscomprehension and criticism, especially when presented in a breezy techno-jargon that provides no proper orientation to readers. This should be separated, however, from the evaluation of the product of that analysis.
There is nothing in the analytic concept of complex-valued covariance that would produce a spurious signal from pure noise as an artifact, as suggested here in irrelevant comments by Spencer and Saumarez. It is the high sampling variability of naive spectral analysis via the raw FFT, not proper low-pass filtering, that produces spectral peaks of no physical significance. And ensemble averaging is not necessary to obtain proper power density estimates from a unique record. At any rate, Vaughan appears to have worked here entirely in the time domain, rather than the frequency domain. A fairly narrow-band ~2.3yr component is likewise evident from proper spectrum analysis of many station records in the tropics and subtropics (very prominently so in the western Mediterranean). It is real; only it’s mechanism is uncertain.
Exploratory analysis of field measurements has a hallowed place in science by providing fodder for explanatory theories. Rather than carping about the fact he doesn’t produce one, it is up to us with training in physics and aspirations of explaining the workings of the climate system to do so.
Paul,
You would be helping readers (and yourself) by providing an introduction that frames the issue, explains why traditional methods of analysis may be inadequate, and encapsulates the analytic distinction between your method and others. There is a relationship here, methinks, to the Hilbert transform that can be exploited for that purpose.
That said, I’m rather mystified why you don’t use ordinary cross-spectrum analysis. There are no serious nonstationarities evident at the high-frequency end that seems to be the focus of your interests, which are just the opposite of mine frequency-wise. Isn’t it cross-spectral coherence that you’re looking for spatially?
Finally, much as I find your results interesting, allow me as an occasional reviewer to point out that your laboratory-notebook jottings do not constitute a publishable presentation.
Cheers nevertheless!
“”””” commieBob says:
May 16, 2011 at 6:06 pm
George E. Smith says:
Well commie, I watched the video several times, and I saw no evidence of any synchronisation.
I saw the video first on http://www.boingboing.net . Here’s a quote.
From the shortest pendulum to the longest, each pendulum is one swing per minute less than the one before it. The effect is a hypnotic dance that repeats every 60 seconds.
In other words, each pendulum has a subharmonic whose period is one minute. “””””
Well commiebob, I don’t deny that the demonstration may have some artistic merit. It’s a toy demonstration for the San Francisco Exploratorium, to amuse children. Edmunds Scientific sells all sorts of “Science” toys, such as molded plastic diffraction grating sheets, and other things to amuse, and attract the attention of inquisitive youngsters.
But the design of the pendula to be one beat per minute different from each other (in sequecnce) is entirely a contrivance, as Dr Svalgaard noted. What would the array do, if the pendula were assembled in a different order, from shortest to longest .
Well of course it wouldn’t make any difference at all, to the end on (distant) view. But provide some coupling from one to the next (by providing a short link from one to another near the top of the strings), and the result would be quite different if the units were rearranged.
Demonstrating that this example, has no correlation whatsoever between the various pendulum components.
I appreciate “Science” art as much as the next person; but it says more about art, than science.
Paul Vaughan says:
May 17, 2011 at 7:30 am
Multiscale complex correlation has not been presented in this article, but it has been measured by 2 different methods.
so what is it?
the dependence of your V reconstruction on space and the skewed distribution of your proxy.
Makes no sense [on space] and the distribution is just the way Mother Nature is.
“A statistician concludes that bigger feet helps the child to learn to read.”
Here you’re just having fun playing to the audience.
no, showing that correlation is not causation.
paraphrasing you:
“In your example it would be technically correct to note a relationship between solar wind changes and climate changes” [if indeed the significance is there which it isn’t]
I explained the filtering above (in response to a question from Erl). I can add that the normalization is based on the maximum absolute deviation.
be more specific. I don’t see any explanation. I had precise questions. Provide precise answers, please.
I am staggered by your statement that one can distinguish whether a signal is random or not by a single record, which you imply from this analysis. I have found this entire thesis unconvincing because it does not appear to convey any new ideas (i.e.: phase aggregation in non-linear systems) and it is completely non-rigorous.
A single observation of a time series does not permit one to make judgements about the distribution of ensemble statistics without very specific restraints.
I suggest that you learn some some proper signal processing before you attempt to take this further (Bendat & Piersol, Papoulis).
Paul Vaughan May 17, 2011 at 7:30 am
“Multiscale complex correlation has not been presented in this article, but it has been measured by 2 different methods.”
Leif Svalgaard May 17, 2011 at 2:51 pm
“so what is it?”
“It” is several multidimensional arrays.
Paul Vaughan says:
May 17, 2011 at 9:50 pm
“It” is several multidimensional arrays.
I was asking what the quantification of the correlation was and you give a non-responsive, meaningless reply.
sky wrote (May 17, 2011 at 10:15 am):
“There is nothing in the analytic concept of complex-valued covariance that would produce a spurious signal from pure noise as an artifact, as suggested here in irrelevant comments by Spencer and Saumarez.”
The myth is pathologically widespread, including in academia. Thanks for weighing in with good sense sky.
sky speculated May 17, 2011 at 10:15 am
“Vaughan appears to have worked here entirely in the time domain, rather than the frequency domain.”
No. What’s above is just a minimalist informal laymen’s sketch.
sky wrote May 17, 2011 at 10:47 am
“That said, I’m rather mystified why you don’t use ordinary cross-spectrum analysis. There are no serious nonstationarities evident at the high-frequency end that seems to be the focus of your interests, which are just the opposite of mine frequency-wise. Isn’t it cross-spectral coherence that you’re looking for spatially?
[…] your laboratory-notebook jottings do not constitute a publishable presentation.”
Publishing formally is absolutely not on my current list of aspirations, but informal discussion of climate patterns is worthwhile. There is temporal nonstationarity in some of the series explored (not all appear above) and part (not all) of it is spatially induced. I have conducted multivariate coherence tests and developed alternative methodologies. Proper elaboration would demand far more time than I can spare at present.
sky wrote May 17, 2011 at 10:15 am
“Exploratory analysis of field measurements has a hallowed place in science by providing fodder for explanatory theories. Rather than carping about the fact he doesn’t produce one, it is up to us with training in physics and aspirations of explaining the workings of the climate system to do so.”
Well said. I also appreciated seeing Erl Happ make similar comments above. Plenty of work to share around on these fascinating puzzles. Thanks for contributing positively.
Leif Svalgaard on his V reconstruction – May 17, 2011 at 2:51 pm
“Makes no sense [on space]”
Would the measurements be the same no matter where the stations were located?
Leif Svalgaard May 17, 2011 at 9:58 pm
“I was asking what the quantification of the correlation was and you give a non-responsive, meaningless reply.”
Some misunderstanding here. It’s not a single number, but rather several arrays of numbers.
Leif, as explained to Erl: Contrast years with immediately-adjacent years.
Paul Vaughan says:
May 17, 2011 at 10:47 pm
“Makes no sense [on space]“
Would the measurements be the same no matter where the stations were located?
The solar wind reconstruction is based on geomagnetic activity which is a global phenomenon calibrated with spacecraft data taken just in front of the Earth [seen from the Sun]. There is no ‘spatial’ aspect to this, and your ‘analysis’ does not include any spatial considerations.
Paul Vaughan says:
May 17, 2011 at 10:53 pm
Some misunderstanding here. It’s not a single number, but rather several arrays of numbers.
Statistical significance of a finding is nor several arrays of numbers. In any event you don’t give any numbers.
Paul Vaughan says:
May 17, 2011 at 10:54 pm
Leif, as explained to Erl: Contrast years with immediately-adjacent years.
I asked how the data was filtered and processed and what spatiotemperal procedure was used. You are ducking everything by not being specific. What does ‘contrast’ mean? Is that a standard statistical concept?
Leif, surely you’re not trying to suggest spacecraft were taking the measurements in 1902.
Paul Vaughan says:
May 17, 2011 at 11:33 pm
Leif, surely you’re not trying to suggest spacecraft were taking the measurements in 1902.
The response of the Earth to solar wind speed is calibrated comparing the response 1963-2011 with spacecraft data.
Leif, a contrast is simply a difference. The contrast of b & c is b minus c.
1. Do you have a plot of the residuals from your V reconstruction?
2. Have you run careful diagnostics?
3. Are you suggesting that the reconstruction is not a function of the location of the antipodal stations measuring aa index?
Paul Vaughan says:
May 18, 2011 at 6:42 am
Leif, a contrast is simply a difference. The contrast of b & c is b minus c.
If so, then simply say it, but you are ducking the question, because what you plotted is plainly not just the difference.
1. Do you have a plot of the residuals from your V reconstruction?
Compare observed and reconstructed: http://www.leif.org/research/Solar-Wind-Speed-1965-2011.png
2. Have you run careful diagnostics?
The only one that counts is how well the reconstruction matches observations [see above]
3. Are you suggesting that the reconstruction is not a function of the location of the antipodal stations measuring aa index?
The aa-index is not used for this. But even if it was, the location would not be important: example: assume the response at station A is twice that of station B: resp(A) = 2 resp(B). Solar wind speed is calibrated against A: V(A) = kA resp(A) and against B: V(B) = kB resp(b). You’ll find that kA is 1/2 kB, because V(A) = V(B). Here is how the speed is reconstructed: http://www.leif.org/research/2007JA012437.pdf
See e.g. Figure 19.
The point is that the solar wind speed measured just in front of the Earth [which is that the Earth reacts to] is a point measurement and has no spatial attributes. You simply get a time series.
Leif Svalgaard wrote (May 18, 2011 at 7:32 am) “[…] what you plotted is plainly not just the difference.”
You’re not reading carefully. There are 2 immediately-adjacent years (the one before & the one after).
Leif Svalgaard wrote (May 18, 2011 at 7:32 am):
“The only one that counts is how well the reconstruction matches observations […]”
There are diagnostics that need to be done on residuals. Do you have a plot of the residuals? What diagnostics have you done?
Leif, there are a few misunderstandings at play here.
First of all, I believe you’re assuming that I imagine iV’ to have variability of a similar spatial form to that of the terrestrial indices. Secondly, you seem reluctant to put clear & direct focus on the fact that early measurements weren’t taken from space. I accept responsibility for saying “aa” where I should have referred to IDV & IHV, but the point is that whether we’re talking about aa, IDV, IHV, or whatever, the early measurements were not taken in space. Also, you’re not being explicit that V does have spatial variability, even if you don’t believe it affects Earth (because it’s somewhere else allegedly having no affect on anything else reaching Earth).
Regardless of whether iV’ has spatial variability that affects Earth, there’s structured temporal coherence with terrestrial oscillations. You’re not suggesting that Earth’s spatial variability can’t respond to temporally-varying spatially-uniform external changes are you?
In the past you’ve [at times] been helpful in sorting out physical possibilities. I suggest that you not be so quick to make false assumptions about what you *think people are pushing or claiming. I’m interested only in the truth. If there’s “coincidental” coherence, I can’t accept that as “coincidental” until I fully understand exactly how & why, via separate causation chains, the 2 phenomena appear synchronized. If you &/or others can’t explain sufficiently, the sensible option is to delay judgement and present interim intuition as speculation.
Paul Vaughan says:
May 18, 2011 at 7:43 pm
You’re not reading carefully. There are 2 immediately-adjacent years (the one before & the one after).
One cannot read your utterings carefully [they fall apart then]. What you plot is not what you say. So, again: what PRECISELY are you plotting? And where did you get the data from? My published yearly V-values from Table 4 of http://www.leif.org/research/2007JA012437.pdf do not match what you plot:
http://www.leif.org/research/iV-prime.png
Paul Vaughan says:
May 18, 2011 at 7:51 pm
There are diagnostics that need to be done on residuals. Do you have a plot of the residuals? What diagnostics have you done?
You may safely assume that I always do a satisfactory job on this. Here is a plot of the residuals normalized to the observed values [from a physical point of view we expect the residuals to scale with the value (i.e. relative error to be about the same)]:
http://www.leif.org/research/Solar-Wind-Speed-1965-2011-Residuals.png
The insert is a histogram of the deviations from zero. There are two areas [marked by ovals] where the residuals are too positive [leading to the slight asymmetry of the histogram]. These represent systematic errors that are known. We left the errors in, because it is not clear how to remove them. See paragraph 38 of http://www.leif.org/research/2007JA012437.pdf for discussion of this.
No further diagnostics are needed as any further errors or discrepancies are systematic errors that one just has to live with. The proof of the pudding is simply how well we reproduce the observed values.
Paul Vaughan says:
May 18, 2011 at 8:55 pm
Leif, there are a few misunderstandings at play here.
It is a bit sad that you in almost any comment lament that there are misunderstandings. Misunderstandings can be avoided by being precise [which is not hard – for most people].
First of all, I believe you’re assuming that I imagine iV’ to have variability of a similar spatial form to that of the terrestrial indices.
‘spatial form’ is IMHO just nonsense [but since you have specified what that means, perhaps I should ask (again) what ‘spatial’ aspects you mean].
Secondly, you seem reluctant to put clear & direct focus on the fact that early measurements weren’t taken from space.
The correct way of comparing datasets is to use data from the same source. Splicing space data and inferred data together is a sin [Mann’s Nature trick] if you do not mark them as different. In http://www.leif.org/research/iV-prime.png I, of course do not the space data at all.
Also, you’re not being explicit that V does have spatial variability, even if you don’t believe it affects Earth (because it’s somewhere else allegedly having no affect on anything else reaching Earth).
What V does on the backside of the Sun is of no interest, and in any case, nowhere in your ‘analysis’ do you use of refer to specific spatial variability, so ‘spatial’ is inappropriate.
Regardless of whether iV’ has spatial variability that affects Earth, there’s structured temporal coherence with terrestrial oscillations.
No, sometimes they are in phase and sometimes they are out of phase as any two random datasets will be.
You’re not suggesting that Earth’s spatial variability can’t respond to temporally-varying spatially-uniform external changes are you?
That statement makes no sense at all. Perhaps you simply mean that the effects might be different in different places of the Earth. Sometimes it has this effect, sometimes that effect, sometimes no effect.
I asked you to consider the possibility that the solar wind observed at Earth depends on the time of year [which would then be a confounding factor], but you ducked that one.
In the past you’ve [at times] been helpful in sorting out physical possibilities. I suggest that you not be so quick to make false assumptions about what you *think people are pushing or claiming. I’m interested only in the truth. If there’s “coincidental” coherence, I can’t accept that as “coincidental” until I fully understand exactly how & why, via separate causation chains, the 2 phenomena appear synchronized. If you &/or others can’t explain sufficiently, the sensible option is to delay judgement and present interim intuition as speculation.
Paul Vaughan says:
May 18, 2011 at 8:55 pm
Also, you’re not being explicit that V does have spatial variability, even if you don’t believe it affects Earth (because it’s somewhere else allegedly having no affect on anything else reaching Earth).
The values of V that I give are measured at the Earth, and by the Earth, so no spatial variability somewhere else enters the picture.
In the past you’ve [at times] been helpful in sorting out physical possibilities. I suggest that you not be so quick to make false assumptions about what you *think people are pushing or claiming. I’m interested only in the truth. If there’s “coincidental” coherence
Making a claim removes the ‘false assumption’ bit. ‘The Truth’? Everybody is pushing their truth. The argument is perhaps that you falsely [to use one of your favorite words] claim there is coherence over and above what we would expect by chance or by confounding with a lurking variable. There are standard statistical methods to measure significance of coherence. Use them.
Leif Svalgaard wrote (May 18, 2011 at 10:43 pm):
“There are standard statistical methods to measure significance of coherence. Use them.”
The whole point is that the assumptions underpinning standard statistical inference are untenable given the nature of the variability & relations. Methods that are suitable for the actual nature of the data need to be developed before more meaningful inference can be conducted.
Leif Svalgaard wrote (May 18, 2011 at 10:31 pm)
“[…] sometimes they are in phase and sometimes they are out of phase as any two random datasets will be.”
We’re not dealing with “random datasets”. See sky’s comments above. The temporal evolution of the coherence indicates nonrandom structure, suggesting that a more physical decomposition is needed to conduct meaningful statistical inference. I get the clear impression that you are lax about performing diagnostics to assess “standard” statistical model assumptions.
Leif Svalgaard wrote (May 18, 2011 at 10:31 pm)
“I asked you to consider the possibility that the solar wind observed at Earth depends on the time of year [which would then be a confounding factor], but you ducked that one.”
On the contrary, I commented directly & clearly; perhaps you overlooked the comment.
As you well know from your experience with the V reconstruction 22 year residuals pattern, the structure of the time series makes sensible decomposition tricky, but a better decomposition is needed to get past untenable statistical model assumptions and towards meaningful p-values & confidence intervals. There is no shame in admitting this Leif; quite the contrary.
I imagine you might find Piers Corbyn’s recent comments entertaining:
http://wattsupwiththat.com/2011/05/17/new-study-links-cosmic-rays-to-aerosolscloud-formation-via-solar-magnetic-activity-modulation/#comment-663406
The temporal structure I find in the coherence suggests to me that a decomposition method conditioned on solar cycle (& perhaps also Hale cycle) phase is needed. As it is, the pattern I get is nearly cyclostationary (not to be confused with stationary). V is a very interesting time series.
Paul Vaughan says:
May 19, 2011 at 7:30 am
The whole point is that the assumptions underpinning standard statistical inference are untenable given the nature of the variability & relations. Methods that are suitable for the actual nature of the data need to be developed before more meaningful inference can be conducted.
Yet you declare significance already.
I get the clear impression that you are lax about performing diagnostics to assess “standard” statistical model assumptions.
The reconstruction of V is not a ‘statistical inference’. We know the physics and the cause and effect chain: http://www.leif.org/research/Physics-based%20Long-term%20Geomagnetic%20Indices.pdf . What other diagnostic would you suggest?
“I asked you to consider the possibility that the solar wind observed at Earth depends on the time of year [which would then be a confounding factor], but you ducked that one.”
You are still ducking this one.
get past untenable statistical model assumptions and towards meaningful p-values & confidence intervals.
The reconstruction is not a statistical inference, but a direct measurement using the Earth as the instrument through a well-understood physical cause-effect.
I imagine you might find Piers Corbyn’s recent comments entertaining
No, silly.
The temporal structure I find in the coherence
You have not demonstrated [nor quantified it] any coherence.
V is a very interesting time series
You are not responding to any of my questions as to the provenance and treatment of V. I showed you that you are not plotting my V.
It’s comical that people have a tendency to suspect some nefariously-motivated, complicated filtering method.
No matter how many times I have to explain it, it won’t get any more complicated:
“i” is a SIMPLE contrast of years with immediately-adjacent years.
That’s it.
Leif posted:
http://www.leif.org/research/iV-prime.png
Looks like you’ve finally got it.
You appear to have used a slightly different smoothing strategy on iV’ (perhaps one not based on the dominant temporal modes in the series?…), but I’ve literally no time to squander splitting hairs — let’s instead simply review sky’s comment that trumps the need for time-squandering hair-splitting:
sky wrote (May 17, 2011 at 10:15 am):
“There is nothing in the analytic concept of complex-valued covariance that would produce a spurious signal from pure noise as an artifact, as suggested here in irrelevant comments by Spencer and Saumarez.”
FYI:
I use RAW data in analyses.
What is posted above is not an analysis, but rather a collection of graphs designed to help instigate a review of serious fundamental WUWT audience misconceptions about the nature of interannual terrestrial oscillations. (Like rubbing a kitten’s nose in its piddle or putting a cat in a bath to promote a cleanly home, one expects to get scratched.)
Even leaving aside the whole iV’ thing & the associated exchange on untenable statistical model assumptions that has Leif bristling, we see from his FIRST paragraph in his FIRST post upthread (the 2nd comment in the thread) that Leif agrees FULLY with the thrust about the relations existing among ALL of the other variables:
Leif Svalgaard wrote (May 15, 2011 at 5:41 pm):
“That all the terrestrial relations are correlated in complicated ways simply show that they are not independent, which we would not expect in the first place. Heaping more of these on top of what is already there will make no difference at all because the inter-dependencies are still there.”
Mission accomplished.
Having to post this article was like having to take out the trash. Just something unpleasant that had to be done.
Best Regards to All.
Paul Vaughan says:
May 19, 2011 at 7:30 am
I get the clear impression that you are lax about performing diagnostics to assess “standard” statistical model assumptions.
Explain what diagnostics [if any] you have performed on your iV’ and iNPI’ correlation.
Paul Vaughan says:
May 19, 2011 at 8:31 am
No matter how many times I have to explain it, it won’t get any more complicated:
“i” is a SIMPLE contrast of years with immediately-adjacent years.
You said: “There are 2 immediately-adjacent years (the one before & the one after).”. So take a year, e.g. 1942. the year before is 1941 and the year after is 1943, so you say you calculate the difference between 1943 and 1941 [careful reading]. Now, I don’t think you do this, but that is what you say.
You appear to have used a slightly different smoothing strategy on iV’ (perhaps one not based on the dominant temporal modes in the series?…),
I do not smooth. I’m trying to ferret out what you do. And you are totally non-responsive, so perhaps it is futile [I recall earlier threads where it was also impossible to get any explanations out of you] to keep asking.
“There is nothing in the analytic concept of complex-valued covariance that would produce a spurious signal from pure noise as an artifact, as suggested here in irrelevant comments by Spencer and Saumarez.”
If the two series have a common confounding component, there is. You are ducking the issue again.
I use RAW data in analyses.
Yet you claim to use a ‘different’ smoothing
Leif agrees FULLY with the thrust about the relations existing among ALL of the other variables
These are all interrelated to begin with and thus bring nothing to the discussion and are just red herrings.
Having to post this article was like having to take out the trash. Just something unpleasant that had to be done.
Looks more like dumping the trash here.
I’m a volunteer Leif. Gravy train riders don’t dictate the terms under which I submit to interrogation. Cheers.
Paul Vaughan says:
May 19, 2011 at 8:06 pm
I’m a volunteer Leif. Gravy train riders don’t dictate the terms under which I submit to interrogation. Cheers.
It is called to chicken out. After having been found out. [snip – play nice ~ac]
http://rcada.ncu.edu.tw/
“About RCADA
Data analysis is indispensable to every science and engineering endeavor, but it always plays the second fiddle to the subject area. The existing methods of data analysis either the probability theory or the spectral analysis are all developed by mathematicians or based on their rigorous rules. In pursue of the rigorous, we are forced to make idealized assumptions and live in a pseudo-real linear and stationary world. But the world we live in is neither stationary nor linear. For example, spectral analysis is synonymous with the Fourier based analysis. As Fourier spectrum can only give meaningful interpretation to linear and stationary process, its application to data from nonlinear and nonstationary processes is problematical. And probability distributions can only represent global properties, which imply homogeneity (or stationarity) in the population. As scientific research getting increasingly sophistic, the inadequacy is become glaringly obvious. The only alternative is to break away from these limitations; we should let data speak for themselves so that the results could reveal the full range of consequence of nonlinearity and nonstationarity. To do so, we need new paradigm of data analysis methodology without a priori basis to fully accommodating the variations of the underlying driving mechanisms.”
Paul Vaughan says:
May 20, 2011 at 7:32 am
To do so, we need new paradigm of data analysis methodology without a priori basis to fully accommodating the variations of the underlying driving mechanisms
Physical scientists do not have a ‘priori basis’ and no new paradigm is needed as the data themselves always have the last word. Of much greater importance is honesty, reproducibility, and disclosure of methods, data, and algorithms. You could try to approach those goals a bit more instead of refusing to explain what you are claiming.
http://rcada.ncu.edu.tw/6%20Instantaneous%20Frequency.ppt
Leif Svalgaard wrote (May 20, 2011 at 7:59 am):
“Physical scientists do not have a ‘priori basis’ […]”
The author you’re quoting [ http://rcada.ncu.edu.tw/ ] was commenting about mathematical scientists, not physical scientists.
Paul Vaughan says:
May 21, 2011 at 8:22 am
The author you’re quoting [ http://rcada.ncu.edu.tw/ ] was commenting about mathematical scientists, not physical scientists.
It was not my quote, but yours. There are no such critters as mathematical scientists. There may be mathematicians or statisticians deluding themselves into believing they are doing science.
Every time a scientist performs statistical inference, that scientist is making assumptions. The author [ http://rcada.ncu.edu.tw/ ] is drawing attention to untenable statistical model assumptions used by scientists (including physicists) to generate misleading p-values & confidence intervals.
Paul Vaughan says:
May 21, 2011 at 9:25 am
drawing attention to untenable statistical model assumptions used by scientists (including physicists) to generate misleading p-values & confidence intervals.
No scientist is dumb enough to believe in p-values or the like. These are only indicative not deciding. The statistical model assumptions are not relevant, the physical model is. Physicists do not make ‘statistical inferences’, but try to compare observations with predictions based on physical laws or hypotheses about proposed mechanisms.
Leif Svalgaard asserted (May 21, 2011 at 9:41 am):
“No scientist is dumb enough to believe in p-values or the like.”
Are you sure about that?
Leif Svalgaard continued (May 21, 2011 at 9:41 am):
“These are only indicative not deciding.”
They’re not even indicative if base assumptions are fundamentally flawed.
Paul Vaughan says:
May 21, 2011 at 11:38 am
“No scientist is dumb enough to believe in p-values or the like.”
Are you sure about that?
None that I know of [and I know a lot]]. Perhaps you know some dumber ones?
Leif Svalgaard continued (May 21, 2011 at 9:41 am):
“These are only indicative not deciding.”
They’re not even indicative if base assumptions are fundamentally flawed.
Nobody has base assumptions that are fundamentally flawed a priori. We judge the base assumptions by the agreement with observations. Or rather, don’t bother with assumptions if the predictions come out even half right.
Leif Svalgaard says: May 21, 2011 at 8:43 am
There may be mathematicians or statisticians deluding themselves into believing they are doing science.
Well said Leif!!!!!
Leif, you’re again misinterpreting the term “assumptions” as if it’s referring to a physical model (& I expected you would do this).
To clarify yet again:
I’m referring to the statistical models upon which statistical inference is based. ALL of these models are founded on ABSTRACT assumptions, REGARDLESS of who is using them.
As for your other misrepresentations & misunderstandings: They are too numerous to address.
Paul Vaughan says:
May 21, 2011 at 2:00 pm
I’m referring to the statistical models upon which statistical inference is based. ALL of these models are founded on ABSTRACT assumptions, REGARDLESS of who is using them.
No, only if the user cares about and expresses abstract model assumptions. Usually he does not, because he knows that statistics [no matter how abstract] doesn’t prove a thing. Can at best give an indication or vague quantification of a relationship that must be established by physical understanding. In economics, statistics reigns because there is no underlying theory or laws.
Leif, like many people, you appear unaware that all statistical inference is based on abstract model assumptions. You also seem to misunderstand that I am an advocate of statistical inference. On the absolute contrary, I am sternly cautioning AGAINST it in contexts where its “reasoning” is based on untenable assumptions (and to be honest I’m not much of a fan of it any context).
Paul Vaughan says:
May 21, 2011 at 4:44 pm
Leif, like many people, you appear unaware that all statistical inference is based on abstract model assumptions.
I’m trying to tell you that physicists do not rely on or base their work on statistical inferences, but on physical theory and observation. A feeling for how little value statistics has you might consult http://www.leif.org/research/Reply%20to%20Lockwood%20IDV%20Comment.pdf
especially paragraph [16].
economics has the law of supply and demand and theories such as the net propensity to consume. newtons law of gravity acts infinity fast at infine distance. newton recognized the physical impossibility.
in science the concept of “understanding” is an illusion. the more we discover, the more questions are raised. we continue to underestimate the size of infinity. like ether, we invent dark matter and dark energy to fill the void.
the value of science is not in “understanding”, it is in our ability to predict accurately, no matter how this is accomplished. If history is a guide, 500 years from now everything we think we “understand” about the physical world will have long since been dumped into the rubbish bin.
ferd berple says:
May 21, 2011 at 5:21 pm
economics has the law of supply and demand and theories such as the net propensity to consume.
Yet their predictions fail spectacularly.
in science the concept of “understanding” is an illusion. the more we discover, the more questions are raised.
It takes understanding to raise more questions.
Leif,
Upthread you were demanding inference.
Now you’re trashing it.
Seems you change stripes as convenient.
–
ferd berple wrote (May 21, 2011 at 5:21 pm)
“the value of science is not in “understanding”, it is in our ability to predict accurately […]”
No interest in predicting, only in exploring. If others are interested in predicting, no problem with that.
ferd berple wrote (May 21, 2011 at 5:21 pm)
“the more we discover, the more questions are raised.”
Exactly.
Paul Vaughan says:
May 22, 2011 at 9:50 am
Seems you change stripes as convenient.
Seems you have left reasonable discourse behind and resorting to person characterizations.
Leif Svalgaard wrote (May 22, 2011 at 9:59 am)
“Seems you have left reasonable discourse behind and resorting to person characterizations.”
On the contrary. Upthread you demanded statistical inference. Now you are saying statistical inference isn’t useful. I have objectively noted the inconsistency.
Paul Vaughan says:
May 22, 2011 at 12:06 pm
On the contrary. Upthread you demanded statistical inference.
On the contrary, perhaps show where I demanded that. I said that physicists don’t do such inferences.
Leif Svalgaard wrote (May 22, 2011 at 12:22 pm)
“[…] physicists don’t do such inferences.”
Nonsense. You’re now suggesting physicists calculate neither p-values nor confidence intervals?
Paul Vaughan says:
May 22, 2011 at 12:59 pm
Nonsense. You’re now suggesting physicists calculate neither p-values nor confidence intervals?
You didn’t show where “I demanded statistical inference”…
And of course, we calculate significance values, but we don’t believe the prove anything. As I said they might be indicative, that’s all.
Paul Vaughan wrote (May 17, 2011 at 7:30 am)
“The annual & semi-annual signals (phase-locked to the terrestrial year) are on my radar […]
Leif, twice you accused me of not saying this. Now more of this ‘he said, she said’ stuff (on stat inference). There’s no value in being ‘tied up at committee’ on such impractical frivolities.
Paul Vaughan says:
May 22, 2011 at 12:06 pm
Now [sic] more of this ‘he said, she said’ stuff (on stat inference).
What happened to: “I have objectively noted the inconsistency.”
Show where.
“The annual & semi-annual signals (phase-locked to the terrestrial year) are on my radar […] Leif, twice you accused me of not saying this.
No, I have asked you how if the solar wind having a component phase-locked to the terrestrial year [but otherwise random] would create a spurious correlation. No answer, of course.
Leif, I haven’t volunteered to participate in an administrative dispute. That would require $70G per year.
Paul Vaughan says:
May 22, 2011 at 5:49 pm
Leif, I haven’t volunteered to participate in an administrative dispute.
True to form, you chicken out. You do seem to have volunteered to ad-hom attacks. I guess some people are like that and do it for free.
When someone instead of stating an objection once, twice, or 3 times repeatedly harasses, certainly there does become an objection to the person (you in this case).
http://wattsupwiththat.com/2011/05/22/arctic-cycles-amopdo-corresponds-to-arctic-station-group/#comment-666101
Paul Vaughan says:
May 22, 2011 at 9:41 pm
When someone instead of stating an objection once, twice, or 3 times repeatedly harasses
And trying to get answers to legitimate questions has now become harassment…
A way of dealing with that is to be responsive rather than abusive (you in this case).
Asking once, twice, or 3 times is acceptable, but please respect that volunteers set the terms on which they choose to respond or not respond.
http://wattsupwiththat.com/2011/05/22/arctic-cycles-amopdo-corresponds-to-arctic-station-group/#comment-666101
Paul Vaughan says:
May 23, 2011 at 7:39 am
Asking once, twice, or 3 times is acceptable, but please respect that volunteers set the terms on which they choose to respond or not respond.
The honorable thing to do would be to acknowledge the request and then perhaps respond with one of
1) I retract what I said
2) I cannot give a meaningful response
3) I don’t give a rat’s ass
The sensible thing to do is agree to disagree efficiently.
http://wattsupwiththat.com/2011/05/22/arctic-cycles-amopdo-corresponds-to-arctic-station-group/#comment-666101
Paul Vaughan says:
May 23, 2011 at 9:36 am
The sensible thing to do is agree to disagree efficiently.
We are not talking about disagreements, but about clarifications. Not wanting to clarify is not sensible, nor honorable.
ferd berple says:
May 21, 2011 at 5:21 pm
the value of science is not in “understanding”, it is in our ability to predict accurately, no matter how this is accomplished. If history is a guide, 500 years from now everything we think we “understand” about the physical world will have long since been dumped into the rubbish bin.
Far be it from me to get into an argument with someone who can’t spell check his own name, but I profoundly disagree with your characterisation of what the value of science is.
While it’s true that our current theory will be replaced in due course, this does not mean that working on and thinking about the concepts which underpin our current conceptualisation of the universe is a waste of time and effort. It has not only got us to the point where we can predict a great many things, which is a useful ability, though not an end in itself; it has also led us towards a deeper understanding and appreciation of the cosmos, and our own minds and being.
Paul Vaughan’s self set vocational task is to find possible connections between disparate phenomena. This is an activity which precedes theorisation and the discovery and quantification of mechanism, which he humbly leaves to others more practised in those arts. We should commend him for ‘making no claims’ and thank him for the interesting discoveries he has shared.
Thanks tallbloke.
=
What the mainstream appears to have overlooked is very simple:
Solar max interrupts the semi-annual heat pump. The frequency of pump outages controls multidecadal oscillations (via hydrology). Interannual spatiotemporal chaos makes this difficult or impossible to see using linear methods.
=
[ For those needing background info:
1. http://wattsupwiththat.com/2011/04/10/solar-terrestrial-lunisolar-components-of-rate-of-change-of-length-of-day/
2. http://wattsupwiththat.com/2010/12/23/confirmation-of-solar-forcing-of-the-semi-annual-variation-of-length-of-day/ ]
Paul Vaughan says:
May 23, 2011 at 1:28 pm
Solar max interrupts the semi-annual heat pump.
There is no evidence for that. How would that happen?
http://wattsupwiththat.files.wordpress.com/2011/04/vaughn_lod2_fig7.png
http://wattsupwiththat.files.wordpress.com/2010/12/vaughn_lod_fig1a.png
http://wattsupwiththat.files.wordpress.com/2010/12/vaughn_lod_fig1b.png