the Air Vent

Because the world needs another opinion

More Evidence of Tamino’s Folly

Posted by Jeff Id on October 24, 2008

To Chris H, Raven and other posts. I do feel privileged to have the quality of comments I get, really it will keep me on my toes. Whether we disagree or not my point recently is to demonstrate some of the flaws in the math on global warming. Tamino has a good point which I cannot and won’t disagree with. His point is in an upslope in temp, assuming the weather noise we have, the local short term down slopes which WILL occur do not preclude a continued trend up. This post deals with your points.

BTW:I verified his ARMA post today, it seems valid to me.

However, — you knew that was coming.

We do have good instruments, instruments are something I am familiar with. Our instruments point out a downtrend recently, something which cannot be disputed. Grant Tamino argues that his ARMA model demonstrates down trends of a certain length do not preclude an overall uptrend. — Grant is right. No problem. However, downtrends are real and an equally valid point is that they may continue.– unlike Tamino no intent to deceive here.

What I did tonight is to look at the last 3 years of downtrend. I made the plots, and calculated the confidence intervals. This time I used actual full noise data rather than the instrument noise for confidence interval calcs.

There full weather noise, only three years and a noise confidence interval of 95%. How can this be argued. We’ll find out.

Oh, yeah I almost forgot. Here’s the code. You will still need to fix the quotes. The CSV file is 30 years of monthly data in 3 columns. If you request my temp data file I can see your email and will send it.

t=read.csv(“c:/agw/giss data/giss rss uah 30yr.csv”, header=FALSE)

x = (1:36)/12+2005

a=gls(y ~x)
plot(x,y,type=”l”,ylim=c(-1,1),xlab=”year”, ylab = “Temp Deg C”)
abline(lsfit(x,y,NULL), col=”red”)
tit=paste(“GISS Slope = “,round(z,4), “, 95% Full weather CONF +/-“,round(w,3))
title( tit)

a=gls(y ~x)
plot(x,y,type=”l”,ylim=c(-1,1),xlab=”year”, ylab = “Temp Deg C”)
abline(lsfit(x,y,NULL), col=”red”)
tit=paste(“RSS Slope = “,round(z,4), “, 95% Full Weather CONF +/-“,round(w,3))

a=gls(y ~x)
plot(x,y,type=”l”,ylim=c(-1,1),xlab=”year”, ylab = “Temp Deg C”)
abline(lsfit(x,y,NULL), col=”red”)
tit=paste(“UAH Slope = “,round(z,4), “, 95% Full Weather CONF+/-“,round(w,3))
title( tit)

15 Responses to “More Evidence of Tamino’s Folly”

  1. FrancisT said

    Are you aware of the graphing site? It looks to me like what you are doing here is somewhat replicable there e.g. this graph with the GISS data from 2004 and trends from 2004, 2005 and 2006 to present

    Yesterday I posted at my blog ( )some more multi trend graphs using the same site and I expect I’m going to work on that a bit more. I find the WFT interface a little annoying but it lends itself very much to automation via perl or Javascript so I’ll probably see if I can make a front end that pleases me more.

  2. Luis Dias said

    rant is right. No problem. However, downtrends are real and an equally valid point is that they may continue.– unlike Tamino no intent to deceive here.

    … Ok. So you found three years with a downward trend. How that proves a “folly” of Tamino remains a big mystery, because, as you admitted, his ARMA model predicts that events like this will always happen in an upward trend.

    So you can postulate whatever you want about the future temps, but I thought the purpose of this post was to prove that Tamino’s post was a “folly”. Thus, epic fail.

    You’d improve your “point” if you found that in the last 30 years such a trend is “unprecedented”, for example, or if you calculated how improbable this event really is in ARMA model (and it would be quite improbable), or how many years of declining temperatures would one need to drive the ARMA model towards 95% improbability (SD1).

    You have done none of this. This post is noise. Greets.

  3. Jeff Id said


    Swingin and missin again. No crystal ball here Louis. I can’t see the future either.

    I found that downslopes exist everywhere in recent data, not just the last 3 years. ARMA data is a model, its not real, the data I used is real, no random numbers here. It would be cool to show trends in an ARMA model, but it would be completely non-useful when looking at how certain we are that temps have dropped.

    The significance is that if someone tells you that there is a drop in temps even in the last 3 years, you have to agree we know it from the data. We even know it to a high confidence even including the (incorrect) weather noise. Something I did just for you.

    Just thought I would say it again. By the way, how did 30 years become such a “significant” amount of time. It’s ridiculous.

    Tamino Foster is a left wing idealist who believes we can legislate climate. His motives are all over the internet. How he was able to confuse a bunch of smart people with a noise model that has downtrend’s is really amazing. Epecially when he crams it in a bunch of dogma blaming the “denialists” for deliberately confusing the issue. It’s a real pile.

    The fact that the extremist AGW movement can’t accept even a short downtrend, “when it’s real” should ring some alarm bells. The fact that Grant Tamino claims it isn’t real at all because “it can’t be measured” should make everyone look close at his motives. Random numbers do not change the data.

    We did measure temp and even with all the GISS station corrections, temps are still going down. Some day they will go up again. Whether we want it or not.

  4. Henry Galt said

    Thanks Jeff. Gotta love it just for the educational aspects.

    If you scroll down 2/3 here:

    to a post by “weatherspout” you will find a graph of a 47 year downward trend in the Hadley, GISS and NOAA products from 1934 to 1980.

    All the while CO2 levels were rising steadily.

    Extremely inconvenient 😉

  5. PaulM said

    A whole website could be dedicated to Tamino’s follies, if anybody had the time and could be bothered to set it up and point them all out.
    In his latest RSS and UAH thread he has difficulty with Fourier analysis, which is odd since he boasts about having written a paper about it (thereby giving up anonymity, if anyone cares). He has a Fourier analysis of a 5 year period, but his frequencies go below 1/5 – in other words his Fourier analysis of a 5-year period contains information about how the series behaves on timescales longer than 5 years! How did he get those numbers? – using his Magic Flute presumably.
    Also he draws conclusions from a 5-year period, having argued in the previous Bjorn Lomborg post that a 10-year period is too short!
    In his ‘Update’ he says that some people weren’t convinced by his ‘amplitude spectrum’, so instead he draws a ‘power spectrum’ – this looks much more convincing if you don’t know (and Tamino doesn’t tell you) that this is just exactly the same graph, with the data squared before plotting it (this is classic Tamino, distorting and stretching the graphs, he does this all the time).
    In his post ‘Spencers Folly’, July 28, he shows that he does not even understand the meaning of the word stable.
    And let’s not forget the recent Ian Joliffe misrepresentation incident…

    Your post above is reminiscent of the ‘consistent with’ series by Pielke at Prometheus – you could look those up if you’re not familiar with them, for example

  6. Diatribical Idiot said

    #2, regarding improbability of recent trend lines…

    I have just updated my HadCrut spreadsheets and charts and will be hopefully getting that post up today. The current non-warming trend line on the HadCrut data goes back to April 1997, which is 11 years and 6 months.

    The last time this long of a non-warming trend line emerged in the data was the 11.5 year period ending June 1967.

    This doesn’t prove or disprove anything, but if the CI around a trend line is as large as Tamino suggests, one would expect more frequent occurrences, I would think. (I’m musing here. I haven’t attempted to prove that thought through some kind of statistical exercise).

  7. ed said

    Do you have a degree in statistics and what in your experience provided the basis for your analysis and commentary on global warming issues?

  8. Bill in Vigo said

    Not wanting to be to bold but where I live all you have to have to Analise and comment on a subject is a desire to do so. IT is called freedom.
    And it sure beats censorship any day. If you disagree present your points and wait for his defense.
    Most credentials used in the study of climate are in other fields of study that over time have become related to the subject. And this is a popular subject and quite interesting for us laymen.

    Bill Derryberry

  9. Jeff Id said


    I am an aeronautical engineer with a professional background in optics, electronics, fluids, structures and computer programming. I have published in optics and spend most of my time running my company. I did my first trend fits using a variety of algorithms including least squares as above more than 20 years ago. I am 40 and I have absolutely nothing to gain personally by my blog.

    In the hundreds of papers I have read, I have never run across anything like I have seen from climatology. Sure the data is noisy, but the continued reinvention of statistics is without question a deliberate complication of an otherwise simplistic field created by people with MOTIVE. I recommend everyone look at the data and the calcs because there is a huge effort underway to pull the wool over our eyes. The hockey stick posts I made are very strong evidence. If anyone tried to pull localized CPS and selective sorting in optics it wouldn’t make it through the door.

    I am not saying man made global warming is false but a high percentage of the science sure as hell is.

    Unlike on Grant Tamino’s blog where I can’t comment, Luis above is welcome to tell me how full of it I am as much as he wants. He can expect an argument though, hopefully a constructive one.

    Tamino Foster is welcome to defend himself here too. But he can’t — because he’s wrong.

  10. hswiseman said

    Jeff, how would you propose determining the threshold length of trend period that constituted “climate” as opposed to “weather” or “noise”? Is Tamino arguing that any trend (up or down) whose length is within the ARMA CI is irrelevant as noise? How long would the trend have to be to become relevant relative to ARMA?

  11. Jeff Id said

    I think Tamino’s method is reasonable. Take slopes through a pile of arma noise. Find SD of the slopes and calculate probabilities. The length would depend on the slope.

    I think it’s pretty amazing that the AGW guys have picked a 30 year trend as climate but an 11 year trend as weather.

    It might be a fun post. I’ll think about it.

    Here’s how I think of Tamino’s argument, I’m sure you have your own thoughts on this as well.

    Grant Tamino fit red noise to a line and said look some negative slopes. He conveniently fails to mention that if the underlying lower freq function is non-linear, i.e. a sine wave or triangle wave which first goes up to 1998 and then down, the probability of measuring negative temps on the way down is higher.

    I think the correct analysis of Tamino’s post is not to say “it’s noise, ignore it” but rather that “temps went down, there is a probability it will turn around again to follow my assumed AGW linear upslope” Each year that goes by which is colder diminishes the probability of a trend reversal within his predicted noise. His claim that we don’t know temps dropped because of a random series on his computer is a joke.

    BTW: ARMA creates a zero mean noise trend. The upslope In Tamiono’s case is a line added to the data. If the linear model fails next year or the one after, all he has to do is modify the underlying linear function and give it some shape to continue his argument in the face of years of downtrend’s.

  12. John F. Pittman said

    Jeff #11. You make a good point with your last paragraph. I think that is also the bottom line in what is being argued in Santer et al in an indirect and non-stated way. If one can’t say with 95% confidence that it is rejected, but it is rejected at the 90% level, the confidence interval would improve by changing the slope (modifying the linear function). I think this poses a real problem with AGW as proposed. They have oversold the 2C/century to the point that any tweaking to the linear frunction as the data becomes available will be seen as a retreat. I would see tweaking as a good and robust scientific correction considering just how few years of data we have for such a noisy (their claim) signal.

  13. Jeff Id said


    I’ve been trying to replicated Tamino’s freq response. I’m getting something similar but for some reason my frequencies are shifted slightly higher. I still think his must be right, maybe something I’m doing wrong in R although I can’t imagine what. As far as the lowest frequencies, the data is sampled to 1/12 year so that will explain it.

  14. ed said

    Thanks for stating your background. My interest was not to tear down what you are doing. I noticed your posts at Realclimate and Climateaudit and could not understand your being blackballed at Realclimate. I’ve had a negative experience myself while trying to edit and add information to Wikipedia that I would consider more AGW “neutral” than the “propaganda” that is presented there. Unfortunately, there is a WIKI editor who patrols anything relative to climate and will not allow any changes or updates that do not pass his evaluation. Thanks and I look forward to reading your future posts.

  15. Cool site, love the info.

Leave a Reply

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

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

Google photo

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

Twitter picture

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

Facebook photo

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

Connecting to %s

%d bloggers like this: