2011-06-11 20:23:45Phil Jones significant temp rise - ask for help & stats question!
Robin2
Robin Webster
robin@carbonbrief...
109.149.104.238

Hi all

This is Robin from Carbon Brief in the UK, with a request for help! Some of you may have seen that the BBC reported an interview with Phil Jones in which he said that acc. to CRU figures, temperature rise since 1995 is now significant to the 95% level. We 've written it up at Carbon Brief - see here http://www.carbonbrief.org/blog/2011/06/global-warming-since-1995-statistically-significant. 

The reason I 'm asking for help is two-fold - first as a young website, we are still building up our community of commentators, and over the last 15 hours or so that thread has been bombarded with skeptics (many of whom I suspect have nowhere else to go to write about this as the BBC post is closed to comments). We could do with some more scientific literacy to redress the balance. So if any of you fancied heading over there for a bit, I would be very grateful.

Secondly, a specific question on stats. A guy called Doug Keenan has  'rebutted ' the Jones statement, saying that the trend is NOT signifiicant using the methodology in the 2007 IPCC report (http://www.ipcc.ch/publications_and_data/ar4/wg1/en/ch3sappendix-3-a.html). He says:

\"Following is an R session showing the statistical calculations. The temperature data (HadCRUT3) was downloaded from the CRU web site today.> t9510<- ts(c(0.275, 0.137, 0.352, 0.548, 0.297, 0.271, 0.408, 0.465, 0.475, 0.447, 0.482, 0.425, 0.402, 0.325, 0.443, 0.476), start=1995)> library(nlme)> confint(gls(t9510 ~ time(t9510), cor=corARMA(p=1,q=0)))                    2.5 %     97.5 %(Intercept) -48.929568004 4.37230179time(t9510)  -0.001989347 0.02462824As shown, the 95%-confidence interval for the slope of the line includes 0. Hence the trend is not significant.Jun 10, 2011 at 7:06 PM | Douglas J. Keenan\"

It looks to me that he has tried to get over this tricky problem by assessing it at the 97.5% confidence level and saying that he is assessing it at the 95% level! However I am by no means a statistician so I can 't feel 100% sure. If anyone out there has this expertise, and is able to post a rebuttal to Keenan on our site, then I think that would be a useful intervention, and again much appreciated.

I won 't have much chance to check comments on this thread today unfortunately as out and about but will be able to let comments through. Many thanks

Robin

2011-06-12 22:05:36
Riccardo

riccardoreitano@tiscali...
93.147.82.178

The definition of statistically significant is somewhat arbitrary. First you have to define the threshold (usually 95%), then you may or may not add other effects like autocorrelation. I won't bother to check what Keenan did, it may also be correct. It doesn't matter.
What Phil Jones did I don't know but I dare say that although he's a great scientist he's not a good communicator. He fell in the trap a few months ago and still can't manage this kind of question.
He has to possible answers. One (my favourite) is to dismiss the question as irrelevant, you can always find a short enough period where there's no statistically significant trend and it doesn't mean you can say nothing; good practice requires that you extend the length of the data (which we have) to assess the trend. The other is to say that even if not statistically significant still it is our best guess, the most probable trend; the chance of having no trend is just a few percent anyways.
I'd not push this story of the short term trend much, it may easily backfire.´╗┐

2011-06-13 13:08:19Reframing the question
John Cook

john@skepticalscience...
121.222.9.229

I recommend not getting bogged down in technical, statistical issues as that just adds to the confusion. Then the public come away with a "he said/she said" impression and come away more confused than ever.

The most effective way to debunk misinformation is to provide an alternative narrative, a compelling story. In this case, the story is the denier tactic of trying to trap a scientist with a gotchya question, the climate equivalent of "do you still beat your wife?". Dana is writing up a blog post on this topic:

http://www.skepticalscience.com/thread.php?t=1867&r=11

For the record, I concur with Riccardo about the poor job Phil Jones has done. You can excuse his confusing answer when he was first ambushed with the question. But a year later, he's still playing into the denier's hands. The issue isn't whether the 16 year trend has pipped the 95% threshhold or not. The issue is that deniers took a short-term trend with a 93% chance of warming and tried to sell it to the public as NO chance of warming.