Lewandowsky’s most recent blog post really makes one wonder about the qualifications at the University of West Anglia Western Australia.
Lewandowsky commenced his post as follows:
The science of statistics is all about differentiating signal from noise. This exercise is far from trivial: Although there is enough computing power in today’s laptops to churn out very sophisticated analyses, it is easily overlooked that data analysis is also a cognitive activity.
Numerical skills alone are often insufficient to understand a data set—indeed, number-crunching ability that’s unaccompanied by informed judgment can often do more harm than good.
This fact frequently becomes apparent in the climate arena, where the ability to use pivot tables in Excel or to do a simple linear regressions is often over-interpreted as deep statistical competence.
I mostly agree with this part of Lewandowsky’s comment, though I would not characterize statistics as merely “differentiating signal from noise”. In respect to his comment about regarding the ability to do a linear regression as deep competence, I presume that he was thinking here of his cousin institute, the University of East Anglia (UEA), where, in a Climategate email, Phil Jones was baffled as to how to calculate a linear trend on his own – with or without Excel. At Phil Jones’ UEA, someone who could carry out a linear regression must have seemed like a deity. Perhaps the situation is similar at Lewandowsky’s UWA. However, this is obviously not the case at Climate Audit, where many readers are accomplished and professional statisticians.
Actually, I’d be inclined to take Lewandowsky’s comment even further – adding that the ability to insert data into canned factor analysis or SEM algorithms (without understanding the mathematics of the underlying programs) is often “over-interpreted as deep statistical competence” – here Lewandowsky should look in the mirror.
Lewandowsky continued:
Two related problems and misconceptions appear to be pervasive: first, blog analysts have failed to differentiate between signal and noise, and second, no one who has toyed with our data has thus far exhibited any knowledge of the crucial notion of a latent construct or latent variable.
In today’s post, I’m going to comment on Lewandowsky’s first claim, while disputing his second claim. (Principal components, a frequent topic at this blog, are a form of latent variable analysis. Factor analysis is somewhat different but related algorithm. Anyone familiar with principal components – as many CA readers are by now – can readily grasp the style of algorithm, though not necessarily sharing Lewandowsky’s apparent reification.)
In respect to “signal vs noise”, Lewandowsky continued:
We use the item in our title, viz. that NASA faked the moon landing, for illustration. Several commentators have argued that the title was misleading because if one only considers level X of climate “skepticism” and level Y of moon endorsement, then there were none or only very few data points in that cell in the Excel spreadsheet.
Perhaps.
But that is drilling into the noise and ignoring the signal.
The signal turns out to be there and it is quite unambiguous: computing a Pearson correlation across all data points between the moon-landing item and HIV denial reveals a correlation of -.25. Likewise, for lung cancer, the correlation is -.23. Both are highly significant at p < .0000…0001 (the exact value is 10 -16, which is another way of saying that the probability of those correlations arising by chance is infinitesimally small).
These paragraphs are about as wrongheaded as anything you’ll ever read.
I agree that a simple “Pearson correlation” between CYMoon and CauseHIV in Lewandowsky’s dataset is -0.25. However, Lewandowsky is COMPLETELY wrong in his suggestion that this “signal” can be separated from outliers. In the Lewandowsky dataset, there were two respondents that purported to believe in CYMoon and disagree with CauseHIV (both were in Tom Curtis’ group of two super-scammers). I’ll show that these two superscammers make major contributions to the supposed “correlation”. Like Lewandowsky, I don’t believe that these two respondents are present “by chance”: I believe that they are present as intentionally fraudulent responses.
First, the correlation can be replicated trivially as follows:
cor(lew$CYMoon, lew$CauseHIV) #[1] -0.2547965
Second, p~+ 10^-16 can be replicated by diagnostics from an OLS regression of CYMoon against CauseHIV (standardized) as shown below:
ols=lm(CYMoon~CauseHIV,data=data.frame(scale(lew[,c("CYMoon","CauseHIV")]) ))
summary(ols)
Coefficients:
Estimate Std. Error t value Pr(>|t|)
(Intercept) 2.574e-17 2.859e-02 9e-16 1
CauseHIV -2.548e-01 2.860e-02 -8.908 <2e-16 ***
However, Lewandowsky is absolutely off-base in his assertion that the examination of outliers is inappropriate statistical analysis. In fact, exactly the opposite is the case: proper statistical analysis REQUIRES the examination of outliers. Furthermore, in this case, the examination of a contingency table (pivot table) is not only relevant but essential to the examination of outliers.
Examination of diagnostics by a competent statistician requires more than looking at the p-value. Part of any such analysis is examination of the qqnorm-plot for the residuals: this is the second graphic in the standard plot in R. Here are the results for CYMoon~CauseHIV (standardized), a graphic that shows severe non-normality of the residuals. (The dashed blue line shows the pattern from normal distribution of residuals.)
Figure 1. QQnorm- plot for CYMoon~CauseHIV.
A second basic diagnostic is examination for outliers using Cook’s distance: this is the fourth graphic in the standard plot in R. This identifies two points (889,963) as very high leverage:
Figure 2. Cooks’ distance: CYMoon~CauseHIV.
Now, let’s do the contingency deprecated by Lewandowsky, a calculation which shows that there are only two respondents purporting to disagree on CauseHIV and to agree on CYMoon.
with(lew,table(CYMoon,CauseHIV))
CauseHIV
CYMoon 1 2 3 4
1 8 5 116 938
2 1 0 33 34
3 1 0 2 1
4 1 0 2 3
These two respondents are the two respondents identified as outliers from the standard diagnostic (889, 963). Both are already familiar to us as super-scammers who claimed to believe in every conspiracy.
To show just that a “significant” correlation can depend as few as two outliers, I’m now going to simplify the contingency table by considering only two classes: disagree – 0 and agree-1, yielding the contingency table below: two respondents in the extreme, with 14 respondents purporting to only dispute CauseHIV and 8 respondents purporting to endorse only CYMoon, as shown below:
Data=twoclass(lew)[,c("CYMoon","CauseHIV")]
with(Data,table(CYMoon,CauseHIV))
CauseHIV
CYMoon 0 1
0 14 1121
1 2 8
The (Pearson) correlation calculated in the same way as Lewandowsky is -0.1488. I’m now going to show that the two outliers dominate this calculation. (The calculation with a 4×4 matrix is structurally identical but adding up to -0.25.)
r=cor(Data$CYMoon,Data$CauseHIV); r # -0.1487561
There are only four unique points (0,0), (0,1), (1,0) and (1,1) in the contingency table. In the calculation below, I show the contribution of each point to the correlation coefficient. The column headed normdot is the product of (x-mean(x))*(y-mean(y)) divided by sd(x)* sd(y)* (N-1), where N is the number of respondents (1145).
N=nrow(Data) Stat= data.frame(CYMoon=c(0,1,0,1),CauseHIV=c(0,0,1,1),count=c( with(Data,table(CYMoon,CauseHIV)) )) m=apply(Data,2,mean);m Stat$dot= (Stat$CYMoon-m[1])*(Stat$CauseHIV-m[2]) Stat$normdot= (Stat$CYMoon-m[1])*(Stat$CauseHIV-m[2])/(sd(Data$CYMoon)*sd(Data$CauseHIV))/(N-1) Stat$normsum= Stat$normdot*Stat$count
The sum of the normsum column gives the correlation coefficient.
sum(Stat$normsum) # -0.1487561
The table calculated above therefore shows the relative contribution of each point to the correlation coefficient as shown below.
Stat[,c(1:4,6)]
CYMoon CauseHIV count dot normsum
1 0 0 14 0.0086115825 0.009640767
2 1 0 2 -0.9774146183 -0.156318155
3 0 1 1121 -0.0001220419 -0.010939947
4 1 1 8 0.0138517572 0.008861259
___________
Total -0.1487561
One can readily see that the two super-scammers (889, 963) contribute essentially 100% (over 100%) actually of the negative correlation between CauseHIV and CYMoon in this calculation.
Next here is the result of applying the same methodology to the 4×4 contingency table in Lewandowsky’s analysis shown here in order of decreasing contribution to the negative correlation. As above, sum(Stat$normsum) is equal to the correlation.
About half of the negative correlation comes from the 33 respondents who disagree with the Moon conspiracy and agree with CauseHIV (without strongly agreeing).
The other half of the negative correlation comes from seven outliers which contribute -0.138 (about 50% of the correlation), with the two superscammers identified above being the largest contributors. (The other 5 outliers need to be examined individually.)
There is a negative contribution from the 938 respondents who strongly agreed with HIV and strongly disagreed with CYMoon: this seems puzzling at first. What happens is that the centroid is moved off dead center. This contribution is offset relatively by positive contributions from on-axis results (CYMoon – strongly disagree or CauseHIV – strongly agree) : this seems to be fairly characteristic in this sort of sparse contingency table heavily weighted on-axis.
Stat[order(Stat$normsum),]
CYMoon CauseHIV count dot normdot normsum
2 3 33 -0.761 -0.004 -0.142
4 1 1 -8.254 -0.047 -0.047
3 1 1 -5.425 -0.031 -0.031
4 3 2 -2.418 -0.014 -0.027
3 3 2 -1.590 -0.009 -0.018
2 1 1 -2.597 -0.015 -0.015
1 4 938 -0.014 0.000 -0.075
2 2 0 -1.679 -0.010 0.000
3 2 0 -3.508 -0.020 0.000
4 2 0 -5.336 -0.030 0.000
3 4 1 0.328 0.002 0.002
1 2 5 0.150 0.001 0.004
4 4 3 0.499 0.003 0.009
1 1 8 0.232 0.001 0.011
2 4 34 0.157 0.001 0.030
1 3 116 0.068 0.000 0.045
Thus the “unambiguous” negative correlation between CYMoon and CauseHIV arises from the following two phenomena: about half of the -.254 comes from only seven outliers, with the two superscammers contributing the most. The other half is contributed from people who neither endorse the CYMoon conspiracy or dispute CauseHIV.
The results for CauseSmoke are very similar. The negative correlation is -0.236. A little less than half is contributed by only four outliers, especially the two (fake) outliers who purport to both strongly believe in CYMoon and disbelieve CauseSmoke. The balance is contributed from those people who hold plausible views, but did not express that they did so strongly.
CYMoon CauseSmoke count dot normdot normsum
2 3 33 -0.754 -0.005 -0.149
4 1 2 -8.231 -0.049 -0.099
4 3 1 -2.395 -0.014 -0.014
3 3 1 -1.575 -0.009 -0.009
1 4 916 -0.015 0.000 -0.081
2 1 0 -2.589 -0.015 0.000
3 1 0 -5.410 -0.032 0.000
2 2 0 -1.671 -0.010 0.000
3 2 0 -3.492 -0.021 0.000
4 2 0 -5.313 -0.032 0.000
1 2 5 0.149 0.001 0.004
1 1 4 0.232 0.001 0.006
3 4 3 0.343 0.002 0.006
4 4 3 0.522 0.003 0.009
2 4 35 0.164 0.001 0.034
1 3 142 0.067 0.000 0.057
Far from the examination of contingency tables being irrelevant to the analysis, they are essential to it.
The “signal” from Lewandowsky’s analysis is also “unambiguous”: that, using his own words, “number-crunching ability that’s unaccompanied by informed judgment can often do more harm than good”. A thesis that his own work amply illustrates.
Update: Jeff Id asked about the effect of robust regression. I’m working on a longer post on robust regression, but will preview this with the result here. R has a very handy robust regression function rlm in the same style as lm, the default option is Huber’s robust regression. The “robust” correlation between CYMoon and CauseHIV is the robust regression coefficient between standardized versions of each series: the robust correlation is 0.000000 (not Lewandowsky’s -0.254). Lewandowsky’s “unambiguous” result is unambiguous dreck.
fm=rlm(CYMoon~CauseHIV,data=data.frame(scale(lew[,c("CYMoon","CauseHIV")]) ))
summary(fm)
Value Std. Error t value
(Intercept) -2.433000e-01 0.000000e+00 -2.138241e+09
CauseHIV 0.000000e+00 0.000000e+00 -2.938290e+05
Residual standard error: 5.487e-09 on 1143 degrees of freedom
Climate Audit 
131 Comments
This is hilarious. In a sad way. Steve, I know just enough about statistics to know when to shut up and learn. Thank you for the education.
May I say also that Lewandowsky himself makes the point for using “cognition” when examining data and using statistics? I don’t have to know R and SEM et al. to look at what Lewamdowsky has done and claims to see, and think, “This seems wrong.”. I know some science is counter-intuitive…but what Lewandowsky is doing isn’t quantum physics and string theory.
Anyway, IMHO, you are doing good by demolishing his shabby work. A strong step in outwitting your political opponent is to make him appear irrational (any wife whose husband has called her crazy knows the effectiveness of this!). Thank you for your yeoman’s work in demonstrating that we skeptics aren’t yet ready for the cuckoo house.
Bender once wrote, “Just a microcosmic example of the danger of confirmatory pattern seeking, and why we need rigorous check & balances in this business. Especially when the “submit” button is so easy to press.”
Seems peer review in certain trades and the “submit” button represent similar hurdles.
The p-values are computed by lm assuming a normal distribution of residuals. The p-values are so small because the assumptions of the OLS model are not satisfied. And why in the world would anyone think they would be? I may have to “recuse myself” from these threads from now on, they’re making me queasy.
I analyze data for a financial services company and I think if any of my employees did something like this I’d tell them that their future lay elsewhere.
Steve: it’s amazing that Lewandowsky can straightfacedly present lm probabilities from a contingency-table type structure. It’s hard to contemplate worse statistics.
I was going to ask the same question about p-values, when the variables are not normally distributed, and come from a very small discrete set. Can you comment on the appropriateness of using a Pearson’s correlation for such variables? [As opposed to, say, Spearman’s correlation, which seems at first glance to be more appropriate to ordinal data.]
Rank correlation would be better. Pearson correlation is closely related to least squares regression since if I standardize x and y, then regress y on x, the regression coefficient is exactly the Pearson correlation (and the intercept is 0). And least squares regression isn’t appropriate here. I haven’t looked at the data in detail so I don’t know what would be appropriate (if anything).
Steve: I don’t think that that is remotely equal to the problem. The structure is far too non-linear and non-normal. Not even remotely.
Actually what I said was not right, the data are already ranked (from 1 to 4) so rank correlation won’t help in the way it might if we had long-tailed distributions. I continue to think the best way to analyze this data is just to look at the contingency table – which says that agreement with conspiracy theories is rare among the respondent, and not a whole lot more.
It really does look like you are playing squash with a 5 year old.
I’m not one who possesses the expertise to evaluate the data and Lewandowski’s use of it, but it certainly looks like he was the one who “toyed” with it, to use his word, and that you are actually working with it, for what it’s worth.
5 year olds shouldn’t play squash with champions.
In the 20 minutes or so since pressing the Publish button, I’ve added more details using the 4×4 contingency table and some comments on this analysis.
Apologies in advance…
The venerable term “stuffed shirt” comes to mind.
This psychologist is talking about the state of climate science and the withdrawal of “bad” papers. Is he having a pre vision of his own paper’s fate?
Re: ZT (Sep 18 21:43),
You owe me my ability-to-eat-solid-food back!
That seriously has to be one of the worse-acted reads I’ve ever seen. The eyebrow lunges are fascinating. I’ve love to know a body language reader’s take on it. Pontification doesn’t even scratch the surface in describing that performance.
Steve: does he remind you a bit of Eugene Levy playing a pompous professor?
I can see almost any of the old Saturday Live cast in that clip. I can really see John Belushi giving that performance.
Umm, for what it’s worth, Levy was a scion of Canada, from SCTV. As Joe Flaherty and John Candy would say: “blowed up REAL good”
I knew that, but when I looked at the clip, I just saw Dan Akroyd, Chevy Chase, and especially John Belushi flashing before my eyes.
As far as the video is watching his eyes he is reading straight from a queue card or teleprompter. I think this explains his body language and the way he phrases his sentences. He seems to be trying to combat reading, looking at the audience (camera), trying to convey what he is reading in a semi-interesting manure (body language and all), and trying to figure out what he is reading at the same time. Not trying to be mean here, I just noted that even over the course of a single sentence he changes his inference and even more so when going from one sentence to another. I’ve done this when I’ve read out loud before and lost track of the context of what I was reading. So sometimes the sentences sounds like a complete thought, but should have been carried over into the next sentence.
If I didn’t know better, I would have thought that this was a parody.
I guess being filmed for television makes some people act unnatural, but still this video makes me wonder if this guy is for real.
And it isn’t like there haven’t been prior real life examples of people working in areas in which they have no expertise and no formal training (I believe “impostor is the correct term).
So I would naively ask if anybody has actually checked the credentials of Lewandowsky – and Lewandowsky screaming “The nutters are here to see my birth certificates.” to discredit anybody asking about him makes me even more skeptical.
And it wouldn’t have been the first time medical sciences have been the area of choice for impostors.
You could have a field day using his own words to lampoon his own paper – if you could stop laughing long enough!
I’ve watched this video before and was rather surprised it was made by a psychologist. Lewandowsky’s body language is stiff, his vocal inflection sounds contrived and his facial expression is frozen except for revealing micro-expressions.
At the end of the video lewandowsky says: “Science is inherently skeptical and peer review is the instrument by which scientific skepticism is pursued. Circumventing or subverting that process does not do justice to the public’s need for scientific accountability.”
Yet in an article Lewandowsky wrote in 2010 he stated:
“Science carries with it responsibilities such as accountability and subsequent scrutiny—peer review is a spam filter, which works well but not perfectly.
“The true value of a peer reviewed article lies in whether or not it survives scrutiny upon publication.
“If it does not, then peer reviewed science is self-correcting and eventually cleanses the occasional junk that penetrated the spam filter.”
http://www.abc.net.au/unleashed/39148.html
Well his paper has apparently penetrated the spam filter and is being scrutinised now that it has been published on the internet. Will the “self-correcting” system cleanse science of lewandowsky’s junk?
I could never quite reconcile the hectoring, sarcastic bully with the “slim ‘n’ sensitive” Lew pictured on his WordShaping posts.
Now, THAT’S more like it!
Reefer Madness!
I’m not sure why – perhaps because deaf people need to be warned too about the dangers of ree.. I mean, denialists – but I transcribed it…
This is why I am thankful there is Steve. I have come to know too many of these people for my liking. This complete and utter ignorance, barely concealed by the sheer stupidity that it is taken seriously, is exhausting.
Quick, Dr. L.: Here are two observations:
A = (33,45) and B = (72,56)
What is the correlation between x and y? If you started calculating stuff, you don’t understand correlation.
The answer is 1.
Now, if I add a thousand points around A of the form (33 – a, 45 – b) and another thousand points around B of the form (72 – c, 56 – d) where a, b, c, d are all independent random picks from [-1 , 1], what will happen to the correlation coefficient?
Correlation is arithmetic … Anyone calculate that number.
Understanding whether appropriate to calculate that number in the first place is what the discipline of Statistics is about.
Ben Pile over at BH pointed to this video as well. Lewandowsky explains the connections between climate change denialists and believers in conspiracy theories on the basis of one letter to the editor.
Ok, that is just weird. The first one was bad enough. This one brings to mind some kind of Big Brother recruitment video. Beware, we are watching and reporting on some writer to the editor in Australia and New Zealand.
I have a wacky theory about people who enter the field of psychology. After watching this, my theory isn’t that out of line.
“Lewandowsky explains the connections between climate change denialists and believers in conspiracy theories on the basis of one letter to the editor.”
Did John Cook write that letter, or was it some other warmist pretending to be a scpetic?
“…we have to lower the western climate establishment’s credibility…”
Just play this video to people. Conspiracy accomplished.
Ah yes…peer review.
http://www.nature.com/news/through-the-gaps-1.11427
But no problems in Climate Science. Nope. None.
Can’t quote much of it without using “the word”, but suffice to say, more hand wringing from our superiors on how shocking, shocking I tell you, it is that peer-review has some serious problems, and dag nab it, we sure mean to do something about it.
Here’s a quote from that article that should put the fear of something into Lewandowsky – and those whose work ‘n words he accepts as gospel:
And not a moment too soon!
While I’m here … for more Lew-related entertainment/enlightenment, readers might be interested in APS blogger didn’t get Lewandowsky’s title “joke”
It’s obvious to anyone in the UK that this is one of Ronnie Barker’s best sketches.
I knew it reminded me of somebody. Which at once brings to mind one of the great man’s earliest monologues, as Chancellor of the Exchequor, which included something like this:
Said with all the necessary pomposity but, it has to be said, better comic timing than Dr Lew.
I highlighted the fact a negative correlation existed between those two items even if you remove all respondents who claim to believe the moon landing was faked. It’s nice to see that confirmed, but with much more information.
One thing I find interesting is the response that marked 4 for all conspiracies doesn’t show up. That’s because he also marked 4 for CauseHIV. This means the fakest response of them all is basically irrelevant for this example, a point I hadn’t noticed earlier. That’s why 963, the response I believe is third-most conspiratorial, matters instead. This would likely change if you switched CauseHIV for CauseSmoke, as the two responses inverted their answers for those two.
-ZT
I wanted to see if there was some context for the video and so I clicked on the link. There was no info in that regard, but the first comment wins a prize:
‘Maybe his hair is mussed because he just took off his tin foil hat.’
That with the video AND Steve’s direct, ‘here it is 1,2,3,…’ analysis ending with
“The “signal” from Lewandowsky’s analysis is also “unambiguous”: that, using his own words, “number-crunching ability that’s unaccompanied by informed judgment can often do more harm than good”. A thesis that his own work amply illustrates.”
indeed made for a good day. I thank all parties.
Thanks for yet another stats lesson.
This isn’t that hard to grock and seems to lead well into a robust regression post. Is there such a thing as hyperheteroskadastic?
Actually your posts sent me reading on how answers should be weighted in surveys like this. In practice, it seems like applying temperatures directly to tree rings.
Steve: in Huber’s robust regression methodology, outliers are downweighted in proportion to the residual. Thus large-residual outliers receive very small weight. I’ve added in the results from robust regression – the “robust” correlation is 0.0000000.
His online bio says:
To be fair, you can be a great teacher and a fairly pedestrian practitioner. The teaching aspect is quite different from the doing in many ways. Certainly great statisticians don’t always make great teachers.
There’s a fair few great football coaches out there who were very ordinary players.
I teach statistics (including regression) to High School kids, and the kids seem to think I do a good job. But I know that I am lost at any level above that. What gets me about Lewandowsky is not what he doesn’t know, but what he doesn’t know that he doesn’t know.
Will Rogers (who got it from Mark Twain) —
“It ain’t what we don’t know what gets us in trouble. It’s what we know that ain’t so.”
Pretty much climate science in a nutshell.
You can also win a teaching award and not be much of anything except popular with students…..
This is not to disparage universities seeking to recognize and reward high quality teaching. There is a great need for more emphasis on teaching quality. But any award system is imperfect, and Lewandowsky’s videos do not suggest such promising teaching quality.
In any case, the key point is that Lewandowsky is a deeply flawed intellect and scientist, who may or may not be popular with some clusters of undergraduates. Both great researchers and awful researchers can turn out to be popular in the classroom, depending on other factors.
The more average practitioner can be a better teacher, because their abilities more closely match the abilities of their students.
The truly outstanding practitioner often relies on some inner quality that cannot be taught. They don’t know “how” they connect the dots. Only that they “can” connect the dots.
The rest of us, having seen the dots connected, can then understand why it works. But would never have thought to make the connection otherwise.
I’ve just had a look at that award – Jo Nova has Professor Lewandowsky’s CV up, which indicates it is a faculty teaching award for honours supervision. I guess that means it is for supervision of a postgraduate student, but not one doing full doctoral research (perhaps best expressed as a research supplement to an undergraduate degree) as a check round the UWA teaching and learning awards website (http://www.teachingandlearning.uwa.edu.au/teaching/awards/awards/faculty) shows that you can also get a postgraduate supervision award (see the 2008 list of awards – no list for 2007). There are a few things that stand out here (as someone who has had some experience of academia).
Firstly, nomination is likely to have been by the students concerned – until 2006 the students union was involved in the process, before it apparently became internal to each UWA faculty (and without a UWA logon, we can’t access the faculty processes – which is not suspicious, as most unis have this sort of level of protection on processes). This may mean that some students at least found Professor Lewandowsky a good teacher – although hardly uniquely so as there are three awards in the faculty per year – but cannot be used to tell us anything beyond the fact that his students (or some thereof) rated him as a dissertation supervisor.
Secondly, this does not therefore concern his ability to do statistics, only to convince students that what he was teaching them was the correct approach to problems. It seems improbable that a teaching award assessed the accuracy of his statistical understanding, rather than reflected his ability as a teacher in a statistical (presumably, since he is claimng this – be fun to find out otherwise…) context.
Thirdly, there is no evidence that Professor Lewandowsky actually teaches statistics courses that I can find. All the psychology courses (including research methods) are co-ordinated by other people, as we might expect as Professor Lewandowsky is a Professorial Fellow, and therefore likely to be a pure researcher.
So whilst he is probably (we can’t confirm this) a winner of an award for teaching, we might need a bit more information before we can accept this as a mark of statistical expertise, rather than being a good supervisor in a field which may or may not have involved statistics.
Here is his CV from 1998.
http://web.archive.org/web/19980117191330/http://www.psy.uwa.edu.au/user/lewan/vita.htm
At that time his research was:
What he describes as his applied research in 1998 was:
Watchman, please clarify your use of the term “students (sic, I think?) union.” Is this a union of the students in the sense of a labor union or the common U.S. usage in which a student union often includes a cafeteria, coffee shop, game and study areas?
ThankRG
obo watchman, in case he doesn’t get back here for a bit. In Australia, until recently, all university students had to pay a compulsory fee to what was called the ‘students union’. This covered things like the subsidised cafeteria, a building for student activities, sports facilities and various other services, depending on the university. It also typically covered the student newspaper and an elected body called something like the Students’ Representative Council, which often engaged in political activities. The SRC or whatever it was called sometimes had a formal role in broader university activities – for example, in some places it could nominate members of the governing body or various committees.
Compulsory student unionism was abolished a few years ago.
HTH.
> the “robust” correlation is 0.0000000.
Smaller than AE Dessler’s famous Science paper then?
This has to be a small senior common room wager: Who can publish a paper with a correlation coefficient closest to zero and yet convince others of the value of that correlation?
FWIW, my paper highlighting the sensitivity of the result in that Dessler (2010) Science paper to different datasets and time periods was published recently in Earth System Dynamics, also discussed here: http://troyca.wordpress.com/2012/09/17/my-esd-paper-on-cloud-feedbacks-was-published/
Clearly not a “proper” climate science paper since it only has a single author. 😉
Nice!
I am not a statistician, rather a retired chemical engineer, but a long time ago wrote a Masters Thesis on trying to predict the nature of bubbly two-phase flow using a stochastic approach with successive decimation and spectral analysis of raw data. Lewandowsky’s claim “indeed, number-crunching ability that’s unaccompanied by informed judgment can often do more harm than good.” has an element of truth about it but one, I suggest, he hasn’t realised. Without a simple, rational, logical view of raw data, statistical analysis may well be meaningless. One aspect of the arguments about the statistical significance of temperature variations propounded by AGW supporters appears to ignore the simple fact that the so-called increases are much lower than the likely errors in temperature measurements. In other words the conclusions are buried in what Lewandowsky would call “noise”. When the whole world is “noisy” it is rather hard to draw conclusions about trends.
The other aspect about which I have some concern is the treatment of outliers. Who is to say that an outlier is not a genuine element of the sample? Who, then, is brave enough to discount even one?
Quis custodiet ipsos custodes?
The latent construct of hidden intent
Rotational math loosed the skeptics bent
‘Cept the advocate with foot on the peddle
Tweaked the numbers and bent construct mettle
Please, someone explain to me, if Lewandowsky considers the responses to “Climate skepticism” and “Moon endorsement” noise (as he is quoted as saying above) why is that relationship enshrined in the title of his paper? It reminds me of the old saying: figures don’t lie but liars can figure.
Is Lewandowsky actually purporting to determine what people believe apart from what they actually say they believe buy using statistics? If not, how can he call the responses he cited in the title of his paper noise?
“A verbal theory is “worth” literally hundreds and sometimes thousands of models”
Steve should have taken his course on his way back from the UK.
http://aris.ss.uci.edu/~joachim/ss2012/prerequisites.php
Stephan Lewandowsky (University of Western Australia)
“Stephan will provide the introduction to the program and will address the fundamental question of why computational and mathematical modeling is a valuable and indispensable tool in cognitive science. His first learning goal is the insight that a verbal theory is “worth” literally hundreds and sometimes thousands of models, many of which behave in ways that differ from the intuitive predictions derived from the theory. Stephan will also cover the fundamentals of parameter estimation, introducing concepts such as the error surface, the discrepancy function, local vs. global minima, and so on. He will then introduce maximum likelihood approaches to parameter estimation. He will also coordinate the overall flow of the School and will take a coordinating role in introducing MATLAB to students.”
‘valuable and indispensable tool’? Well, presumably if something is indispensable then it is valuable. If this is an example of the precision of thinking that pervades LewWorld, heaven help us.
I don’t want to seem like a nitpicker, but clear language is inextricably tied to clear thinking. Just like the Orwellian expression where Tom Fuller was told he had ‘recused himself’ when he had no say in the process of removing all his posts, this sort of stuff needs to be pulled up. We are not talking about a community newsletter here.
Indeed, the whole screed reminds me of the worst kind of Managementspeak that is used to justify inflated salaries in the private sector. Lots of fluff and not a lot of substance.
What are Lew’s statistics qualifications btw ?
I teach stats to undergraduates in the physical sciences. Its a subtle business and I still don’t feel I’m a “real” expert. Lew clearly has a lot of belief in his stats ability. I wonder what his background is.
I’m slightly surprised that Lewandowsky – being a supposed expert in statistics in psychology – would choose Pearson’s correlation over, say, something like polychoric correlation for this type of a data set. Of course the input for polychoric correlation is that ol’ Excel pivot table, so such approaches clearly cannot be of any use *cough*.
But as Steve notes, that isn’t really the main issue at play here. The main problem appears to be that in a large-ish database of results, one small subset of sample data are corrupted and the methodology applied causes this small number of corrupted inputs to dominate the final results.
Now, wait. Why does that sound familiar?
Perhaps he has Mike Mann envy.
Stephan Lewandowsky seems a mite coy about the subjects of his BA and MA degrees.
Click to access SLvita.pdf
B.A. Washington College, Chestertown, MD, U.S.A., 1980
M.A. University of Toronto, 1981
Ph.D. University of Toronto (Supervisor: Bennet Murdock), 1985
http://websites.psychology.uwa.edu.au/labs/cogscience/Stephan_Lewandowsky.htm
Brief biographical sketch
I completed my undergraduate studies at Washington College, Chestertown, MD, USA, in 1980. I then did my post-graduate training at the University of Toronto, earning a PhD in 1985. After various stints as a research fellow, I took up my first full-time academic post at the University of Oklahoma in 1990. I moved to the University of Western Australia in 1995, and I have been here ever since.
Steve: hmmm, we’re both alumni of the University of Toronto. Lewandowsky knows or ought to know the difference between math as learned by contest-winning math students and psych students and be a
littlelot more cautious in his public sneering.http://www.psych.utoronto.ca/users/murdock/
I took an invitational competitive math test when entering college due to perfect placement scores and placed in the top 10 of over 5000 incoming freshmen. It was a surprise because I was always a terrible student in HS with no advanced mathematics, so I had no previous recognition of any skill from schools. I finished my engineering math in the first two years of school without difficulty and solved some unusual math problems which stumped engineering professors along the way. Our company is based on solving unique math problems in optics, so it isn’t that small of a compliment when I write that I have learned more math from Steve and others at CA since college than anywhere else. The topics here require me to go back and learn and re-learn which is one of my favorite parts about this blog. Lewie’s mathematical self-immolation while indirectly referring to CA has been rather humorous to watch.
Oddly, Lewandowsky and I are both alumni of the University of Toronto. The math-track students (there were only about 15-20 in pure math in a large university) were in a completely different league than psychology students. We were supposed to understand the concepts that we were working with, not just apply recipes. Lewandowsky should understand the difference between math students and psych students.
I had lunch with one of my classmates a couple of years ago (who’s now a math prof at U of Toronto.) The teaching of statistics within applied fields like psychology by psychologists rather than math/statistics profs has been a point of contention. The solution is not obvious since the math profs tend to get too academic.
But there are clearly pitfalls with someone like Lewandowsky – who apparently doesn’t understand the pitfalls of a ‘significant” correlation.
It looks like Lewandowsky is familair with the issues of outliers. http://websites.psychology.uwa.edu.au/labs/cogscience/documents/Spence-Lewandowsky%20(1989)%20-%20Robust%20MDS.pdf This makes his handling of this data set even more strange.
I went through Eng/Phys at the Univ. of Alberta back in the early 1980’s, when you took 7 or 8 classes per semester. I sat in on several honors math courses at the same time, and noticed a huge increase in level of difficulty compared with the honors physics courses math and/or the EE courses math content, even in upper level courses. I can’t imagine the difference between honors math and psych math or arts math courses.
@bernie1815 thanks and definitely strange. The final sentence is ‘Also, after a robust fit has been obtained, it is often rather easier to identify bad data points via residual analysis than when a conventional procedure has been used.’
Lewandowsky is quite familiar with outliers but is currently trying to obscure the issue.
AT UBC in the mid-60s I was fortunate one year to have as math prof someone who had done applied mathematics at a well known California university.
But later I had the misfortune of of having a “teaching assistant” in a subject that I view as very complex matrices (can’t remember name of course). Probably one of the graduate students given TA jobs so they get some income.
(I did not have Dean Gage as prof, renowned master teacher (probably in early years). Like John Ridpath (economics at York), Gage won teaching awards.
Re: Jeff Condon (Sep 19 09:13), Jeff: I placed in the top 2% — here, I feel I am in the bottom 2%. Like many here I am not that humble, however, there are times when acknowledging superiority exhibits wisdom. My suggestion is simply that Lewandowsky should re-consider his motivations and attitudes if he wishes to advance his field of study. Further research is a definite requirement…
I noted Lewandowsky’s arts (not science) degrees a couple of days ago at WUWT. His failure to disclose the fields of study is interesting.
Perhaps these degrees explain his use of artistic license when using statistics.
Check his list of publications, especially around the time he received his PhD. Based on my cursory look at some of his early papers, it is unwarranted to assume that he is unfamiliar with mathematical statistics. On the other hand he is using least square methods with 4 point scales that have skewed distributions. I would have thought that non-parametric methods would be more appropriate.
I suggest that someone from Oz who is a LinkedIn subscriber take a look at the Lew’s LinkedIn profile. Oz is out of my network. Also, any U of Toronto alums who belong to their alumni association should be able to get degree info, thesis and dissertation titles, etc. from the alumni assoc. or the department.
I think you are far too kind to these authors. The department of psychology to which they are affiliated will certainly offer a statistics course. Now that R has pretty well displaced SPSS, texts like Crawley’s “Statistics”, or something like it, will routinely be used as an introduction. So students will know when linear regression is seriously ill-advised – it’s usually in Chapter 1 and is often asked about in examinations. They will have been taught how to do Q-Q plots to identify and deal with outliers. They will know about Cook’s distance. Students will know about the difficulties inherent in analysing count and categorial data. They will have been taught some measurement theory (Australian psychologists made important contributions to this topic). I don’t know whether they are routinely taught about robust regression, but they will have been taught about related concepts in the context of analysis of variance. And so on … and so on. These issues are all in the standard texts. I simply cannot believe the authors don’t know about them. Indeed, I can’t believe they don’t know a lot more than their students. So they have a much more serious charge to answer.
With you there RA.
One things that amazed me for a long time is how the academic perfromce of some individuals within climate ‘science’ is actual worse than students undertaken degrees when it comes to some very fundamental factors. That your lectures should be unable to meet the standards in their own work they demand of you in yours ,even if just an essay, really should concern any student as to the ‘quality’ of the education your getting .
If they had done the job properly, they’d have had to report that they found nothing of any significance. Which given the size of the grant they got to do the work, wouldn’t look very good.
This is a fundamental point, ignoring which can make us unnecessary conspiracists as far as Lew and team are concerned. But why on earth that size of grant?
Good question. How much do a few pc’s, some paper and pens, a jar of coffee and a free online survey cost? $2k?
Agreed. UWA even has a Center for Applied Statistics, which says it provides extensive support and consulting within and beyond UWA, but a senior Prof such as Lewandowsky may feel he learned what he needed long ago. It would be interesting if some of UWA’s real statisticians could be persuaded to review these matters:
http://www.cas.maths.uwa.edu.au/contact/staff
When you spend decades looking a tables of earth science measurement data, it becomes almost like a language. I worked through the era of manual data gathering – pencil, paper, log book, eraser, graph paper for plotting – through the mechanical calculators, through the wonders of the HP-45 and later calculators, to the first of the affordable benchtops. At the last stage, a lot of field work was avoided while people huddled over desks learning Basic or whatever became available. Then, when we could afford colour printers that drew graphs on roll paper, we became able to contour a set of data without even looking at the figures.
Now, some colleagues and I tried to extend the manual plotting of data for as long as possible, so that people retained the ability to ‘feel’ it. Of course progress marched on and these days very few people do manual graphs or contours.
In one landmark case of a geochemical survey, we found several quite high zinc anomalies in a cluster. Maybe 5 points in 5,000. The first inclination was to grid at closer spacing, but the geochemist at the base was more wiley. He looked up the history books and found that there had been a small plant there to recover gold using zinc as one of the reagents. That’s what we were picking up. He did this because the numbers did not ‘feel’ right.
So, for many purposes, I agree that manhandling the data can produce gains that automatic processing cannot. The converse is the use of automatic processing to produce anolamies or deductions that can be synthetic. That seems to be the case here.
To conclude the story of some traps in the march of progress, we worked vehicles in very remote and dangerous parts of the country, where no water for 2 days = death. When GPS navigation came in, the lesser experienced said “At last we can throw away our air photos and maps and navigate accurately without looking out for landmarks”. Again, we intereceded with two simple consequences. 1. What if the batteries in the GPS failed? 2. What if the US military fritzed the GPS signal accuracy during an exercise or military crisis?
A common thread here is failing to have a planned objective. Sure, science has cases of serendipity, but most of the time it is hard slog work that HAS to be done to an a priori hypothesis. Those who have preconceptions or just follow the data can run into trouble. Yes, I have erred at times as well.
Geoff, while I have worked in very different fields, I know what you mean about ‘feeling’ the data. As Agatha Christie said in her crime novels, what is called intuition is very often just a subconscious bringing together of your past experience and knowledge.
People who spend their lives sitting in offices or labs, with no ‘feel’ for the real world, can make colossal blunders which a first year trainee in the field would avoid.
It’s also known as a BS detector, something which is essential to healthy survival outside academia.
When did “peer review” become the foundation for science and skepticism? I don’t remember any mention of p.r. when I was taught the scientific method back in the Stone Age.
Fascinating comments from an eminent microbiologist on why it is a mistake to “deify” the current peer review system:
http://phylogenomics.blogspot.com/2012/02/stop-deifying-peer-review-of-journal.html
What will it take to remove them (or the next generation like them) from gatekeeping? That depends on precisely what support, or direction, they have from above. I don’t expect it to be easy.
I took the survey just recently. I think it was at WUWT. I can’t remember if I even finished since the wording of the questions was so biased and imprecise.
A simpleton’s view of how a free market works, etc.
On the smoking question, I answered Agree. I did not answer strongly agree since the question used the imprecise term “causes” instead of the scientifically correct “greatly increases risk” or “is strongly correlated with” Since they did not provide the moderation in the question I did so in my answer.
An excess of academic language does not undermine the basic reality that if you put rubbish in you tend to get rubbish out , when you get to the bottom line that is what Lewandowsky did . Everything followers from his need to support the physiology profile he drawn, and has become addicted to , of AGW sceptics as not just wrong but mad or bad. In itself not usual for the hard-core of AGW proponents, its is nevertheless completely wrong and so can only be ‘proved ‘ by the application of even more rubbish .
I enjoy watching an expert play with numbers it makes the hairs on the back of my arms stand up.
Also reminds me that such cognitive reasoning cannot be random, my dog sits when I tell him but he expects a treat in return.
Think I hit GO before inserting the URL
oh dear –
why I guess it is progress that on Prof Lewandowsky’s blog – ‘he who should not be mentioned by name’ – Steve Mcintyre, gets mentioned by name.
Prof Lewandowsky cannot bring himeslf to stop making accusations of motive
– Heading – A simple recipe for the manufacturing of doubt
http://shapingtomorrowsworld.org/oberauerEFA.html
The link title and the changed order of authors of the latest post suggests it may be Oberauer who did the analysis.
I understand Lewandowsky to point out two errors:
He links to Exploratory Factor Analysis
His post doesn’t really respond to anything said in this post, but it’s still worth looking into. For the first claim, it’s easy to redo what Steve did with a different number of factors. Since rotation (I believe) is a non-issue when you only use one factor, the best case to test is the climate items. All you have to do is modify the corresponding line in Steve’s code to say “factors=1.” When I do, I get .827, not the .86 claimed by Lewandowsky’s paper.
That suggests neither the number of factors extracted, nor the rotation used, is enough to explain the discrepancy. If so, that’d mean Lewandowsky’s results are either wrong or affected by some unknown step.
As for the the second claim, it is cheeky to say someone used the wrong rotation scheme if you never bother to disclose which one you use. How is someone supposed to know what decisions an author makes if he doesn’t disclose them? And if Lewandowsky wants to criticize Steve for not using the right rotation scheme, why didn’t he state which scheme was the “correct” one? What’s the point of hiding it?
Thanks Brandon for checking. (PS I should have stated: “Lewandowsky thinks McIntyre made two errors.”)
No problem. Fair warning though, I’m no stats expert. It’s possible there’s some obvious point a person who knows there stuff would catch that I’m missing.
I’d be confused as to why Lewandowsky didn’t point it out though.
One such obvious point would be the difference between “there” and “their.” D’oh.
Brandon, it is unlikely that one can reproduce the Lewandowsky factor analyses using the factanal function. That function uses maximum likelihood as the method for calculating the loadings and other parameters in the fit. If the paper (which appears to have been written 7 or 8 years ago) referred to by David is his earlier comment is any indication, the basic method of choice by Lew and friends is principal axes which is a variation on principal components where the diagonal 1’s in the correlation matrix are replaced by other values. Other criteria have been introduced for doing the calculations as well so the possibilities are quite numerous.
A good R library for this purpose may be psych. Their fa function can do many of the different procedures and in the case of multiple factors, can do a large variety of rotations including several versions of geomin which would probably be used by Lew since it is the default in Mplus factor procedures. For the record, I have not been able to exactlyduplicate any of the results in their paper although I have gotten reasonably close.
Compliments Brandon on your patient dogedness.
RomanM
faustusnotes 17:40 PM on 20 September, 2012 attributes the differences to the program used:
Or faustusnotes 17:48 PM on 20 September, 2012 attributes the differences to:
I would have thought both programs would have been sufficiently validated by now to give the same results for the same input parameters.
RomanM, thanks for that comment. I think it makes sense to use principle axes since the data is nowhere near normally distributed. I just didn’t think that would matter to the authors since they seem comfortable calculating confidence levels which require the assumption of a normal distribution!
Kidding aside, I did consider that, and I actually tried that exact package. I focused on the climate item latent variable since rotation shouldn’t matter for it, and by using the covariance instead of correlation matrix, I was able to get 85% explained variance. This is not the 86% reported by Lewandowsky, and nothing else I tested could push it higher.
I think it makes sense to use a covariance matrix here since the scales are the same for all five items, so that might be part of what the authors did. However, it seems there is something else involved, and I can’t think of what it could be. And if I/we can’t figure out how to replicate the simplest result, I doubt replicating the results where rotation matters will be possible.
Anyway, unless I’m mistaken on the math, it shouldn’t matter whether one uses principle axes factoring or maximum likelihood factoring if you’re only extracting a single factor. That’d mean it doesn’t matter which I use for my current testing.
I’m now able to exactly replicate the puzzling “explained variances”. They do not come from factor analysis whether it be maximum likelihood or principal axes. They come from principal components (?!?). If you do a SVD on the correlation matrix, you get Lewandowsky’s results. I’ve got a detailed post showing this, which should be online in an hour or so.
lewandowsky is quite a gong show.
That’s… incredible. And I don’t mean that in a good way.
I’ve never heard of the term “gong show” before but I like it. The Urban Dictionary defines it as –
“An event marred by confusion, ineptitude and shenanigans”.
GrantB
This is the Gong Show:
Very much like climate science and Lewandowsky’s work; you can dress up a pig…but it’s still a pig in a dress.
😉
We seem to get second rate academics from USA. Others are Will Steffen and David Karoly. None of them understand statistics. Maybe they think they will be more noticed as big fish in a little pond.
The danger of being a big fish in a little pond in Oz is surely that as the summer heat comes you become a bleached fish in a non-existent pond. Not a bad analogy – taking the warming as purely seasonal of course. And McIntyre is applying the heat now.
No, David Karoly is home grown. I worked with his father.
Steve,
The good Professor Lewandowsky says:
Wait a second !?!
The Lewandowsky, et al., study is a typical but poorly designed opinion survey!
Lewandowsky is using it to tie respondents’ opinion matrices on Moon landings, assasination
plots, HIV/AIDS, and backed-up toilets with their failure to believe in the man-made global
climate change syndrome (skeptics). This does not take the study out of the realm
of standard opinion survey statistical practices and jump it into some special “climate arena”
where Manninan statisical techniques flourish.
Any American undergraduate student who passed their Psych Department’s Survey of Psychological
Testing course, or their Political Science Department’s Opinion Survey Design class, and did OK
in either/both Department’s stastitics classes will find flaws in almost every step of this
survey from the skewed universe of respondents down ro the final interpretation of the data.
What are they teaching at those universities down under ?
How to protect the near extinct proper-panda?
I studied polling and analysis at both undergraduate and post graduate level in Australian universities (although not at UWA). Let me assure you that this exercise would have been comprehensively failed in any course I took. In fairness, the good Professor apparently learned his trade in other countries!
Re: johanna (Sep 20 11:12), Ouch! Mr. McIntyre is Canadian — as am I. Low blow — but perhaps deserved. 😉
Sorry, I wasn’t having a go at UT or anywhere else. What stands out to me is the extraordinary disparity between the standards expected of students and the staff who are supposed to be role models in terms of expertise, experience and integrity.
I really hope that Professor Lew’s students are not being taught that what he did is even acceptable, let alone something to aspire to. None of my teachers would have touched this incoherent farrago with a ten foot pole.
I watched the Lewandowsky clip about “deniers” and peer review twice. Absolutely fascinating. I hope that Josh or Minnesotans for Global Warming are able to use this precious, illuminating gem to further elucidate the issue, the problem with climate “science,” today.
Between NOAA attempting to clamp down on PBS for having the temerity to interview Anthony Watts yesterday, to McIntyre today….and this Lewandowsky cut, the mask is slipping. Badly, and fast. What’s underneath is ugly, indeed.
And, sooooooooo sadly stupid! (But, funny!)
Soon, it is going to be impossible to maintain the facade, even among the devotees.
…Lady in Red
Josh has come up with some top art work.
http://bishophill.squarespace.com/blog/2012/9/19/save-your-scientists-josh-185.html
Josh is good!
It’s obvious that some noses are bent and egos badly bruised over at Lew’s World. A tad sad and I might feel sorry for them, if they weren’t so damned arrogant and pompous. Pomposity married to idiocy is always awaiting a banana peel around the next corner.
A little humility would serve them well. (But of course they are all “credentialed experts” — if not educated, intelligent, smart or packed with a smidgeon of common sense.)
Now we need a Hide the Steps song from folk in Minnesota…. What silliness Lew and his gang engage in, refusing to “educate” Steve. …too funny?
Separate the analysis, however, the survey was jr high school juvenile. This is no search for information, knowledge — truth? It’s a smear campaign, paid for with govt research funds…. I would never have completed the survey were it not for Anthony’s request. The questions were absurd, as were the possible choices.
The survey says nothing about me, my politics, my attitudes, even what I know. So who cares if Lew’s analysis is flawed? Why bother? ….Lady in Red
Damned if you don’t, damned if you do. From 2005 in http://www.realclimate.org/index.php/archives/2005/01/on-yet-another-false-claim-by-mcintyre-and-mckitrick/ We already demonstrated the falsehood of this assertion here by showing (a) that the hockey stick pattern emerges using either the MM (centered) or MBH98 (non-centered) PCA conventions, but was censored by MM through an inappropriate application of selection rules for determining the number of Principal Component (PC) to retain, (b) that use of the correct number of PC series (5) to be kept with the MM (centered) convention retains the characteristic “Hockey Stick” pattern as an important predictor, and yields essentially the same temperature reconstruction as MBH98…
From Lewandoski’s latest post: Our EFA of the climate items thus provides clear evidence that a single factor is sufficient to represent the largest part of the variance in the five “climate science” items. Moreover, adding further factors with eigenvalues < 1 is counterproductive because they represent less information than the original individual items. (Remember that all acknowledged standard criteria yield the same conclusions.)
As a receiver of a PhD and MSc in Quensland, Australia I can assure that their levels were well below thos I obtained in South Africa (MSc). So Steve the problem with Australia education is not confined to Western Australia but the whole country. Australian should stick to tourism and mining. LOL
Adrianos, I’m torn between lampooning you for such an outrageous inference on the basis of a sample of one – yourself – and agreeing with you regarding the ‘problem’. Australian universities have been turned into fee-factories. In the good ol’ days of general grants to be disposed of any way the institution saw fit (thereby encouraging a lot more pure research) Australian universities were rigorous institutions, but now it’s all about marketing and PR in pursuit of ever more fee-paying students, and sucking up to the government of the day for targetted grants. But there remain some islands of excellence. My children are at Monash and Melbourne universities doing traditional arts (dead languages, European history, classical literature, music) and the standards and expectations now are in some ways higher, and in no way lesser, to my own experience 30 years ago. Maybe science is different. We used to joke at degree ceremonies about how the science faculties, with fewer students on average than arts, seemed nonetheless to issue PhDs at a ratio of about 30 to 1.
Those that know me would be incredulous to see me say the following (after reading Steve’s statistical breakdown and Lewandowsky’s lame attempt at his own supposed claim of statistical acumen), I am totally speechless.
re: “Lewandowsky should understand the difference between math students and psych students.”
Back in 2004 two Sociology grad students at Berkeley researched with statistical analysis showing the electronic Presidential voting in Florida had problems. Their Sociology Professor endorsed their research in a press briefing. Their analysis had similar statistical shortfalls.
I’m seriously considering the development of a website to serve as a Press Kit for the climate science/IPCC conclusions debate.
The disinformation in the debate is thwarting reasonable dialogue and the Press needs to understand the difference between Alarmism and the Science.
I know from experience, when a logical approach occurs to me its likely already available.
Is anyone aware of a document, site, or effort that has defined the taxonomy of the various self-labeled groups in the climate science debate?
Properly defining the true meaning of these self-labeled terms and their origin would be a logical dance card for the Press. It would also serve as a logical tool to evaluate Alarmist generalizations like those we find in the Lewandowsky paper.
I’ve pulled together a general shopping list of terms and was struck by the origin and definition I found for Warmer and LukeWarmer.
see Steven Mosher’s comments:
Making the lukewarmer case
Posted on June 1, 2011
by Judith Curry
Thanks in advance for the help.
Feel free to delete my last comment if its over the top for some reason.
Michael Mann has Lewandowsky’s back.
“Stephen McIntyre goes back to same dishonest “recipe for
manufacturing doubt” behind his attacks on Hockey Stick. His latest attacks on Professor Stephen Lewandowski’s new study on climate change denialism actually exploits the same sleight of hand (throwing out important signals in a dataset) behind his original attacks on the Hockey Stick reconstruction, as detailed in “The Hockey Stick and the Climate Wars” ”
Mann calls McIntyre “dishonest” … wonder if there’ll be a lawsuit?
I mean fair is fair right? I’ll pitch in.
😉
Re: A.Scott (Sep 19 17:33),
An Anthony makes the big time media ‘hit squad’ – Media Matters … who also tread dangerously close to libel with accusing him of lying:
A clear implication that Anthony was “paid by the Heartland Institute to sow doubt about climate science” as well.
These people are off their bloomin’ rockers – Anthony simply presented a different view – did so rationally and straightforwardly – even acknowledging climate has warmed.
So now Mann with his all “noise” and no “signal” is an expert in the principles of sampling theory as well!
This is a primary example of what has been wrong with the world of climate scientists.
Each one is perfectly willing to endorse the work of the other outside their narrow frame of expertise without having two sweet clues about whether the science was correctly done – as long as the result supports the cause. No need to admit mistakes or apologize for incompetent academic work because the others will defend you without question.
Sometimes it makes one feel like some Greek guy constantly rolling a large rock up a hill. Sisyphean takes on a whole new meaning…
These guys probably believe the episodes in cop shows where they blow up the image of a license plate 100x and get a perfectly clear “signal”…
And they simply will not accept the fact that sometimes your sample is too small to do anything and sometimes all you have is noise.
@ZT
>> the “robust” correlation is 0.0000000.
>Smaller than AE Dessler’s famous Science paper then?
>This has to be a small senior common room wager: Who can publish a paper with a correlation coefficient closest to zero and yet convince others of the value of that correlation?
I thought that was a pretty good suggestion as a social science experiment on the skill of stats confabulators and the gullibility of publishers.
The problem is that Our Man Lew published a paper with a correlation coefficient of 7 decimal zeroes but claimed one of -0.254 which is cheating just a little bit, wouldn’t you say? Would it not be a better challenge to try to express the zero in an opaque formulaic fashion replete with blather about one having to be intelligent enough to know when 0.0000000 = 0.-254? But that prize is already taken.
Further to the comparison with Mann above, it really does seem as if he is convinced others can’t see through his work, if he did it knowingly. If it was unknowing, as I suspect, I doubt he can follow Steve’s analysis.
Slaughtering Lews bad math is all well but the two worst offences are the extremely poorly framed questions (so bad that often the answer requires lots of assumptions) and the sampling at venues not known to be frequented by sceptics. And then i havent even commented on the fact that he classifies roger pielke jr as a sceptic.
Check mate McIntyre you goose. lolololollolol
http://www.shapingtomorrowsworld.org/oberauerEFA.html
Lew spews:
“Now you know why the title of our paper was “NASA faked the moon landing—Therefore (Climate) Science is a Hoax: An Anatomy of the Motivated Rejection of Science.” We put the “(climate)” in parentheses before “science” because the association between conspiracist ideation and rejection of science was greater for the other sciences than for climate science.”
That is precisely the opposite interpretation that the reader would draw. Calling out “climate” with the parenthetical serves to emphasize the link to that field, not discount it.
This guy has some serious issues.
The problem is not with Lew’s spews but with the cause of them.
Warmists….they are the subjects tested in the study.
Whatever conclusions about skepticism and conspiracy ideation and their veracity, they only apply to the study respondents, ie “believers in CAGW” = “conspiracy theorists”. We know their origin, no matter how they self-identified. Based on their approach to dealing with skeptical viewpoints, Lew’s views might appear to have some foundation.
What must make make being on the wrong end of this beatdown worse is the good-natured professional manner of it. Anyone can deliver a loud, spittle flecked bollocking, and half the time if you yell louder they back-off. This though, that’s gotta hurt… I wonder if the “institution” he “works” at can find a stat’s professor who can attempt an even half convincing reply. I suspect they’ll speed read this post and tell the Spanked Newsworthy Lad that he’s on his own…
Lewandowsky has obviously become accustomed to speaking with students, who he can flummox with his authority. He is so used to this that in this realm he forgets to take a step back, realize who he debating, and actually try to make sense. Good for you Steve for exposing him..
Apparently, no one’s made an announcement, as yet, on Lew’s World. I thought about it but the honor should belong to another. Brandon?
My late husband would get up Sunday mornings and do the NY Times Sunday Crossword in ink. In about 15 minutes. It was just a pleasant exercise for him.
Reverse engineering Lew’s work I suspect was not easy, especially insomuch as it appears the job was not an obvious straight line logical one. I wonder about the effort and manhours Lew invested concocting his analysis stew.
And, then, Steve just reverse engineers the absurdity. Too funny.
….Lady in Red
PS: I still think the whole damn survey is a waste of time, regardless the analysis.
I hope Steve found it enjoyable, like a complicated sudoku puzzle.
The statis-tick deLEWsion?
From Page 13 of Lewandowsky’s paper:
another objection might raise the possibility that our respondents willfully accentuated their replies in order to subvert our presumed intentions. As in most behavioral research, this possibility cannot be ruled out.
However, unless a substantial subset of the more than 1,000 respondents conspired to coordinate their responses, any individual accentuation or provocation would only have injected more noise into our data.
This seems unlikely because subsets of our items have been used in previous laboratory research, and for those subsets, our data did not differ in a meaningful way from published precedent. For example, the online supplemental material shows that responses to the Satisfaction With Life Scale (SWLS; Diener, Emmons, Larsen, & Griffin, 1985) replicated previous research involving the population at large, and the model in Figure 1 exactly replicated the factor structure reported by Lewandowsky et al. (2012) using a sample of pedestrians in a large city.
Lewandowsky, S., Gignac, G. E., & Vaughan, S. (2012). Climate science is not alone: The pivotal role of perceived scientific consensus in acceptance of science. Manuscript submitted for publication.
I am not exactly sure what “exactly replicated the factor structure reported by Lewandowsky et al. (2012)” means, but I think he is saying because he got the about same distribution of answers for a subset of questions as he got from a previous survey, there could not have been a significant number of faked answers.
If this is correct, I am not sure how the answers from one population group can validate the answer from a distinctly different group.
In the case of this study, he studied the subset who hold extreme anti-cagw views and also participate in extreme pro-cagw websites; in other words, his study group -if they were actually real – was compromised almost entirely of ‘internet trolls’; a very different population group than ‘climate sceptics’
6 Trackbacks
[…] the rest here at Lewandowsky’s Fake Correlation GA_googleFillSlot("Wordads_MIS_Mrec_adsafe"); //Rate this:Share […]
[…] That’s an epic fail if I’ve ever seen one. Even Nuccitelli’s buddy, Stephan Lewadowsky’s statistical blundering on his “skeptics deny the moon landing” paper isn’t that bad. Tamino will not be […]
[…] Professor Stephan Lewandowky, whose laughable paper has been so devastatingly critiqued by Steve McIntyre who showed how just two fake “sceptics” answering his dodgy survey seem to have given him his claim that […]
[…] https://climateaudit.org/2012/09/18/lewandowskys-fake-correlation/ […]
[…] issue of false responses received a lot of attention with Stephan Lewandowsky’s paper, “NASA faked the moon landing—Therefore (Climate) Science is […]
[…] That’s about as scientific as saying my cat produces more hairballs in winter due to “climate change” But wait, there’s more! He’s part of the Lewandowsky alternate universe of made up data correlation: […]