Open Thread

Again, I am getting sloppy on publishing these regularly…

Possible topics for discussion:



22 thoughts on “Open Thread

  1. Frank, I don’t see it so much as psych of bullshit, but as an unfortunate trait in psych of humans, that’s exploited full-force by confusionists. So it’s psych of why the bullshit that’s being used is so effective (which is why it’s being used…)

    Hey, has anyone else had experience of late with ostensibly-general-public people sidling up* to you in life, & making odd “casual” statements or queries that sure look like they’re designed to elicit oppo-research info or the semblance thereof? e.g. a fellow train traveler mentioning that her sister “worked as a contributor to SourceWatch”, a concertgoer asking “oh, you’re in the Communications business?”, etc… (well in a general sense yes, since my aim *is* to communicate…)
    *And in some cases sidling up *close*, in a personal-space-invasion way. And other times encouraging you to engage in dubious behavior.

    • Anna:

      Frank, I don’t see it so much as psych of bullshit, but as an unfortunate trait in psych of humans, that’s exploited full-force by confusionists.

      Well, that’s a bit too long-winded — I think I’ll stick to “psychology of bullshit” for now. :)

      Hey, has anyone else had experience of late with ostensibly-general-public people sidling up* to you in life, & making odd “casual” statements or queries that sure look like they’re designed to elicit oppo-research info or the semblance thereof?

      Not me. I’m not sure what to say here except that you probably want to find ways to reduce leakage of your private information, so that people can’t use it for shadowy ends. Anyway, I just found that EFF has a web site on Surveillance Self-Defense, which may be useful…

      — frank

  2. What does it mean that random data is more likely to correlate to temperature data than other random data?
    I calculated that white noise correlates at 10.6% but temperature at 13%.

  3. Carbon taxes, do they really work? Or do they become a massive fraud? It seems to me that all previous attempts at reducing any emissions via a “tax” or “trade” have been monumentally unsuccessful, with billions of dollars lost to the criminal element. To me that is like subsidizing any industry. As long as the taxpayer is making it profitable, there is no real sustained effort made to improve the technology. Remove or reduce the subsidies, and the houses of cards tumble. There has to be a better way.

  4. Frank, here it is.

    double [] temps={-.521,-.219,-.264,-.367,-.004,-.438,-.53,-.412,-.251,-.031,

    double [] random=new double[temps.length];
    double [] random2=new double[temps.length];
    double corrtgt1=.106;
    double corrtgt2=.128;
    int countcorrtgt1=0;
    int countcorrtgt2=0;
    int numruns=100000;
    double [] corr=new double[numruns];
    for(int j=0;j<numruns;j++){
    for(int i=0;icorrtgt1) countcorrtgt1++;
    if(corr[j]>corrtgt2) countcorrtgt2++;

    System.out.printf(“%d %d”,countcorrtgt1, countcorrtgt2);

  5. That code doesn’t look remotely like a correlation coefficient. I have no idea what it’s supposed to calculate.

    For anyone who’s interested in how to do this calculation correctly, I wrote a little R code. (I assumed that -448 is supposed to be -0.048, and changed the uniform random numbers to Gaussian random numbers, as is usual for white noise.)

    Results in R:

    temps = c(-.521, -.219, …, .864)
    N = length(temps)

    cor2 = {r1 = rnorm(N); r2 = rnorm(N); c(cor(r1,temps), cor(r1,r2))} # noise-temperature and noise-noise correlation

    correlations = replicate(100000, cor2())
    rownames(correlations) = c(“noise.temp”, “noise.noise”)

    mean(correlations[“noise.temp”,]) # noise-temperature correlation = -3.63e-6
    mean(correlations[“noise.noise”,]) # noise-noise correlation = 7.73e-5

    As expected, noise-noise correlations are on average zero, and so are noise-temperature correlations.


    sd(correlations[“noise.temp”,]) # std. dev. of correlations = 0.0799
    sd(correlations[“noise.noise”,]) # std. dev. of correlations = 0.0801
    1/sqrt(N) # theoretical std. dev. = 0.0798

  6. Hmm, it appears the blog software is killing the code.
    it’s actually for loop of i from zero to random dot length, random and random2 bracket i are set to the same as the lines above for random bracket zero,
    corr bracket j is set to correlationvalue of temps and random, and if corr bracket j is greater than corrtgt1, increment that by 1.

  7. MikeN,

    Maybe you calculated the correlation coefficient, but the code is missing. You declare a ‘corr’ array but I don’t see any code that assigns values to it. You assign values to the first elements of the ‘random’ and ‘random2’ arrays but to no other elements (maybe random[0] should be random[j]?).

  8. Mike,

    It’s probably a less-than sign in your for loop; that needs to be escaped into an HTML entity.

    Anyway, regardless of what your missing code does, if you’re getting nonzero correlations, on average, then I think there’s a problem. I also don’t understand why you’re counting how many replicates have a correlation above a certain specific threshold, or where those thresholds came from. I thought the question was, does temperature correlate with random white noise? On average, it does not.

    By the way, there’s a typo in my code, replace:

    cor2 = {r1 = rnorm(N); r2 = rnorm(N); c(cor(r1,temps), cor(r1,r2))}


    cor2 = function() {r1 = rnorm(N); r2 = rnorm(N); c(cor(r1,temps), cor(r1,r2))}

  9. Nathan, thanks. My code does have an error in the temperatures. I’ll see how that affects things. This code came about while discussing Mann 08, which tests 1200+ series for a minimum correlation to temps, and I found the result surprising.

    I accept that the average correlations would be 0, but what percentage of correlations are above a certain threshold? The numbers I gave above were for .1 correlation, when I really wanted .106 per Mann’s paper. White noise against white noise correlated above .1 10.6% of the time, while the temp series correlated 13% of the time. I had realized I didn’t use Gaussian distribution, but when I saw the .106 that Mann used in his paper, I decided maybe I did it correctly..

  10. The distributions of correlations should be the same, to within numerical error.

    sum(correlations[“noise.temp”,] > 0.1) / 100000 # 0.106 = 10.6%
    sum(correlations[“noise.noise”,] > 0.1) / 100000 # 0.107 = 10.7%

  11. Yes, the two came out the same once I fixed the temperature error. What do you get for red noise? This is why I had random[0] only assigned in the outer for loop. Using the same random function, I am consistently getting temperatures to be 1% higher correlation matching. Is this to be expected against a growing function?

  12. Martin Clark:

    I see you wrote 6 paragraphs about your supposed climate ‘skepticism’ arising from the urban heat island effect, but I see not a single word of explanation about why the UHI effect would turn a non-warming trend into a warming trend.

    The only way the UHI effect can turn a non-warming trend into a warming trend is if this effect becomes more and more pronounced over time. Not only have you not demonstrated this to be the case, you’ve not even tried to demonstrate this!

    I’m curious to know why you think that your 6 paragraphs of irrelevant information are more useful than a simple calculation (which you didn’t do) showing the effect of UHI over time.

    — frank

    A nice explanation of this common myth is available at Skeptical Science. -Kate

  13. Late to the thread, but…

    FYI, the carbon tax in Australia has been announced, and drafts of the legislation to enable it have been publicly released for comment, but it hasn’t yet been introduced to the parliament for voting.

    Having said that, it’s generally expected to pass – the Labor Party, The Greens, and several independents have said they’ll support it, and the opposition (generally opposed to anything climate-related) don’t have the numbers to block it, although they’ve promised to repeal it if elected in 2013 (although the total package has been cleverly built, so to repeal it, they’ll have to promise to raise taxes on the low & middle income brackets, cut aged & disability pensions, and rely on the ‘goodwill’ of large corporations to lower prices for electricity & other carbon-intensive goods).

    There’s a big PR campaign going on here, though – mining & fossil fuel companies and some conservative business groups are running ads in all the major papers alleging it’ll destroy the economy, drive up the cost of everything, and not make any difference globally.

    The proponents have tried to counter that, sometimes pointing to the EU or the British Columbia examples, but it’s been difficult to get such counter-examples any airtime – the only national daily newspaper has a well-known bias against accurate reporting on climate (see e.g. the lengthy series of posts on Deltoid about “The Australian’s War on Science”), which is assisted by most of the other media reporting the “balance” rather than the “scientific facts”.

    I’ve had some quite scientifically-knowledgeable colleagues bring up some of the basic climate myths as reasons why global warming isn’t real. Sometimes they’re quite shocked when I debunk the myths.

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 )

Facebook photo

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

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.