Daily Mail caught in another lie

Following the heels of the Rosegate scandal where journalist David Rose was exposed as a serial quote fabricator, the credibility of Rose's newspaper, the Daily Mail, has taken another body blow with the paper publishing a false story claiming that Phil Jones had admitted that there had been no global warming since 1995.

This is false (see graph below) and Jones made no such admission. Michael Tobis has the details on the Daily Mail's dishonesty.

i-2d0aa8eb506bc2b59672458359e0d45c-gisstempfiga.png

Ever gullible Tim Blair, of course, swallowed the lie, hook, line and sinker. Andrew Bolt will do doubt follow if he gets his voice back.

Update: Glenn Reynolds and Ann Althouse fall for the lie.

More like this

I wish Jones had resisted the request for an interview; it only results in abuse of his integrity. He seems extraordinarily unwary despite these 'journalists' repeated demonstrations that you wouldn't trust them with your lunch order.

@Nick: he gave the interview, and apparently in writing, with the BBC. The Daily Mail then 'interpreted' the interview. And the Daily Mail has shown itself quite apt at making up things. It was in the forefront of the MMR scare, and now again in stirring up anti-AGW ideas.

If, as people are claiming the Daily Mail have fabricated its reporting, then who is going to take legal proceedings against them?, i find it very strange that only last week Jones was playing the sympathy card with a "contemplating suicide" story, Then the next thing he is on the BBC doing an interview,
Im sure its a very noble thing to do albeit foolhardy as he is under the microscope by the media & his colleagues not to mention the legal investigations ongoing,
The fact still remains that Jones did say there had been no warming as confirmed in the emails which he admits are genuine,
Even the BBC are now questioning AGW which, considering they are the largest government propaganda peddlers are slowly waking up to the fact that the science is not "settled" & it does need closer scrutiny as there is alot of doubts that have not been validated.

By chumney warner Bsc (not verified) on 14 Feb 2010 #permalink

What percentage of people reading that article understand the meaning of significant in a statistical context?

By This really is… (not verified) on 14 Feb 2010 #permalink

@chumney warner Bsc:

You are aware of the costs of legal proceedings? And in case you think you get your money back when you win, ask Ben Goldacre. When he was sued by Mathias Rath he was lucky enough to have the Guardian pay for him. Lucky, I say, because despite Rath losing (or rather, retracting the case), the Guardian spent about 300,000 pound MORE than they got back.

In other words: one person going against an institute with plenty of lawyers, has absolutely no chance.

@3 All that this sad bullshit boils down to is a journalistic culture of shoddy research,unethical misrepresentation and quote-mining,compounded by rapid dispersal. This is what sustains an essentially phony interaction between some scientists and those in fear of their work. Please don't be part of it and put words into Jones' mouth. Just read the BBC interview and don't attempt to paraphrase. And download the IPCC reports,take the time to work through them; that is the only way you'll have a chance of making sense.

Chumney said:

>The fact still remains that Jones did say there had been no warming as confirmed in the emails which he admits are genuine

Incorrect. The emails do not confirm any such thing.
You are imposing your own interpretation on them, there isn't enough context in the emails to suggest what you are suggesting.

@chumney warner Bsc

Are you serious? Anyone who reads the odd scientific paper quickly has to understand what _statistically significant_ actually means. The trend is still positive, But only just "not significant at the 95% significance level". Translation: "There is just under a 95% probability that the gradient of the real life trend is not 0."

If I were Jones, looking at this retrospectively, I'd be kicking myself for phrasing it that way. Instead of saying it isn't "significant at the 95% significance level", he should have put it in more positive, general terms, such as "we can be 94% confident that the planet has continued to warm since 1995". At least that way the Daily Mail couldn't quote mine to get what they want.

"Instead of saying it isn't "significant at the 95% significance level", he should have put it in more positive, general terms, such as "we can be 94% confident that the planet has continued to warm since 1995". At least that way the Daily Mail couldn't quote mine to get what they want."

In that case, the Daily Mail would write that Jones "isn't certain" and "expresses doubt"...

I posted a comment to the daily telegraph article. We should post rebuttles to the comment sections of these articles when we see them. At least that way a casual reader will not just get the huff and fluff from the denialist commentators and will at least be handed the possibility that the article and denialist commentators are wrong. What I wrote follows, I took the numbers by eye from woodfortrees, I wasn't hoping to be 100% correct just throw some sort of balance out there:

Phil Jones did not claim There has been no global warming since 1995.

He said there had been no statistically significant warming since 1995. There is a big difference.

No statistically significant warming since 1995 means the period 1995 onwards has too much noise to determine the whether the trend is positive or flat with statistical significance. So this article, and the daily mail one behind it, is wrong to claim Phil Jone's has said there has been no warming since 1995.

This is why climatologists look at 30 year periods with respect to climate because that's a long enough period to distinguish the trend from the noise with statistical significance.

The linear best fit of hadCRUT since 1995 shows 0.15C warming. That isn't enough warming to be statistically significant over this 15 year period.

Note that another 15 year period, 1980 to 1995, showed an even lower amount of, about 0.12C. So we could have had an article in 1995 claiming no statistically significant warming since 1980. Yet now with more data - the 30 years since 1980 show 0.48C warming, which is indeed statistically significant.

"...the credibility of Rose's newspaper, the Daily Mail, has taken another body blow..."

You must be kidding Tim. The Daily Mail had no credibility to begin with.

> on which the Mail base their article.

Well, in the same sense that the romance story in the Titanic movie is 'based on' historical fact.

@14

From his own mouth it came

It has to be admitted that Jones seems a blithering idiot to use, in a mass-media interview, scientific terminology that's inevitably going to be misunderstood by the general reader.

Even so, Miss Magoo, now this terminology has been explained to you not once but several times on this page, why do you persist in saying "From his own mouth it came"?

If I were Jones, looking at this retrospectively, I'd be kicking myself for phrasing it that way. Instead of saying it isn't "significant at the 95% significance level", he should have put it in more positive, general terms, such as "we can be 94% confident that the planet has continued to warm since 1995". At least that way the Daily Mail couldn't quote mine to get what they want.

Yes, exactly. He did a horrible job in that interview, giving honest, non-controversial answers in a way that provides several opportunities for malicious quote-mining.

Which, of course, is exactly what's being done with that particular statement, and his answer to the question regarding the MWP.

RE 17:

The BBC web site has the following Q&A.
======================

BBC - Do you agree that from January 2002 to the present there has been statistically significant global cooling?

Jones - No. This period is even shorter than 1995-2009. The trend this time is negative (-0.12C per decade), but this trend is not statistically significant.

====================================
.
If Jones had applied your suggested approach to the non-statistically significant trend from 95 to the present, it would have been logical to do the same for the period 2002 to the present.

Would you have wanted Jones to say something like "We can be 94% confident that the world has cooled since 2002"?

@11 Lurker

Ha, yes I stand corrected. The Daily Mail would almost certainly still spin it any way they wish. Phrasing it differently would have eliminated any ambiguity for those who seen the direct quote however.

Although I admire Jones' honesty in telling it exactly how the science tells it, I believe it further highlights the need for scientists working in controversial fields such as this. Sometimes they just need reminding that the general public don't have a scooby what statistical significance means.

A little snippet of my post above got cut off:

That should say - "I believe it further highlights the need for scientists working in controversial fields such as this to have even a little PR training."

Chuck @ 18,

If Jones had applied your suggested approach to the non-statistically significant trend from 95 to the present, it would have been logical to do the same for the period 2002 to the present.

Would you have wanted Jones to say something like "We can be 94% confident that the world has cooled since 2002"?

1995 to 2009 = 15 years. 2002 to 2009 = 8 years. What makes you think both would be just outside the 95% confidence range? Jones says 'only just' to the first one but a flat 'no' to the second - the reason being the shorter time period adding greater uncertainty, of course. I hope you see the logic here.

For a more detailed explanation of how long you need to make conclusions about trends, see this very informative post by Tamino: http://tamino.wordpress.com/2009/12/15/how-long/

Would you have wanted Jones to say something like "We can be 94% confident that the world has cooled since 2002"?

No, because such a statement would be meaningless. Really, so would be the statement about warming from 1005-2009. The discussion is about how to get relevent messages across in a way that won't be spun, not about removing the concept of statistical significance per se.

Anyway, I'm not even sure you can tell the truth and not have it spun. It's an unfortunate side-effect of working in an arena that has political implications.

Actually Jones doesn't need to study PR. Like to many scientists, even scientists in academia, he needs to study pedagogy.

By drunken orangetree (not verified) on 14 Feb 2010 #permalink

The real kicker here is that when someone like Latif actually looks into the issue of short term temperature changes it gets blown up into PROOF that AGW is a giant conspiracy....

What I don't understand is why he didn't have a PR person with him to help him out (or if he did, why he didn't have a good one). No 'celebrity' would be in this situation talking to a paper without an adviser, and I don't see why scientists should be without one. Politicians wouldn't go near the papers/BBC without hours of briefing by their adviser on what to say. This sort of help isn't dishonest; instead it's important that scientists have a chance to say what they mean without their words being twisted.

I suggest you look to see what the London Times has to say today about this

Tim already has.

By Chris O'Neill (not verified) on 14 Feb 2010 #permalink

@woodforthetrees:
Phil Jones *did* have a PR assistant for the interview with the BBC. But you can't do much about a newspaper taking that story and twisting it, other than giving no interviews at all. Which is bound to be twisted, too...("see, he's not rebutting anything, so he lied").

Nobody is listening to you anymore, Lambert. It must be terrible to consistently be on the wrong side of history.

Murph carefully ignores the large number of blog posters, and indeed the links to people such as Lucia who listened to the debate.
Wrong side of history? Is that reification or what?

Anyone sort the Daily Mail article comments by "worst rated"? One was rated down because they suggested it was a good idea to go green even if for no other reason than to get off the oil needle.

And we wonder why the world is going to hell. (in more ways than one.)

Does anyone else wonder if conservatives know there is global warming but are rooting for it anyway? The wild fluctuations in weather and disasters that accompany it are what they want.

It fits their end the world scenario for Jesus' return. Anything they can do rush Jesus along is good.

I think that the case of Phil Jones shows exactly why climate change needs to get effective PR people on board. If the PR person sitting beside him during the BBC interview didn't think that the sort of language he was using might be misinterpreted, then they need to be in a different business.

Does the Daily Mail lie? Yes (My parent still read it every day, so I'm familiar with their MO). Will simply keeping quiet keep you off the front page? No. Will journalists write rubbish (like the Times article), if they can get away with it? Yes.

The only way to stop this stuff, or at least fight back, is to clearly talk to the media and thus the public, and for all of us to bombard the media every time they get it wrong.

And for anyone from the UEA to never go near a journalist or microphone ever again...

Is it wrong for me to think that everyone who understands that there is a growing problem due to AGW should just be quiet for the next 40 years, buy lots of provisions, build green communities in safe areas and, when the water rises, the sandstorms blow, and the average climate hits its projected 5 degree increase, do a hearty, globally directed Nelson Munch from The Simpsons "Ha-ha!"?

Probably, yes. But it would feel good.

This is what Jones said;

" in the last 15 years there had been no âstatistically significantâ warming,"

Considering, in the last 15 years there has been "no âstatistically significantâ warming", yet the carbon levels have increased unabated; how long is it going to take you AGW fraudsters to admit you are wrong?

By Devils Advocate (not verified) on 14 Feb 2010 #permalink

>*how long is it going to take you AGW fraudsters to admit you are wrong?*

How about you show with statistically significant that there has been no warming for 15 years?

Unlike Jones you don't seem to understand [what it means](http://tamino.wordpress.com/2009/12/15/how-long/) in this context.

"You you don't seem to understand what it means"; because it means exactly what it says, which is "in the last 15 years there had been no âstatistically significantâ warming," i.e. "no âstatistically significantâ warming," in the last 15 years.

By Devils Advocate (not verified) on 14 Feb 2010 #permalink

What it doesn't mean is that AGW is false. Try this for a null hypothesis - there is actually no underlying trend for the last 15 years.

This test presumably comes out with something like p=0.06, as it's only just missing that significance level. This is not quite enough to reject the hypothesis that
there's no trend at the 95% confidence level.

You could also try... 'there isn't an underlying trend of +0.15C/decade for the last 15 years'. I don't have the tools to perform that test, but you bet your ass that it comes out with P<<0.05, based on knowledge of the data and the above result. Damn, we can reject that hypothesis then.

What it doesn't mean is that AGW is false. Try this for a null hypothesis - there is actually no underlying trend for the last 15 years.

This test presumably comes out with something like p=0.06, as it's only just missing that significance level. This is not quite enough to reject the hypothesis that
there's no trend at the 95% confidence level.

You could also try... 'there isn't an underlying trend of +0.15C/decade for the last 15 years'. I don't have the tools to perform that test, but you bet your ass that it comes out with P<<0.05, based on knowledge of the data and the above result. Damn, we can reject that hypothesis then.

Devil's Advocate (#37/#39),

Define "statistically significant warming". I dare you.

"Try this for a null hypothesis - there is actually no underlying trend for the last 15 years." If you said that you would be digging yourself into a bigger hole.

It is not what you or I say that counts here; what counts is that Jones has turnaround 180o and said;

"no âstatistically significantâ global warming," so it is fraudulent to say that there has been statistically significantâ global warming over the last 15 years.

Now, how long is it going to take you AGW fraudsters to come clean and admit a fraud has been committed; Another 15 years?

By Devils Advocate (not verified) on 14 Feb 2010 #permalink

>"You you don't seem to understand what it means"; because it means exactly what it says, which is "in the last 15 years there had been no âstatistically significantâ warming*

Which does not mean there has been no warming, and does not mean there should be statistical significance in the warming signal over such a short period when mixed with other cycles and noise.

And it also means your quip about "AGW fraudsters" is not supported by the data nor the quote you provide. Infact your mis-use of the the meaning in this context is evidence of your fraud.

It means,the difference is small enough to be utterly unimportant. Basically, it takes the W out of AGW or it makes the W utterly unimportant.

BTW my other posts are being censored or blocked.

By Devils Advocate (not verified) on 14 Feb 2010 #permalink

Unsurprisingly, DA gets it exactly wrong. Statistical significance has nothing to do with importance. One could hypothetically have a statistically significant trend of .0000001 degrees of warming per decade, an amount which is for all intents and purposes not important. (of course, you'd have to have either a vast time range or data with practically no noise to find this, but hypothetically it's perfectly possible). No statistically significant warming â  no important warming.

Super question MarkB. Delicious Advocaat has clearly never taken a statistics course... or even stuck 'statistical significance' into google.

Too much to hope for, I suppose.

By saying the warming is NOT statistically significant, it is the same as saying the warming is so small (& utterly unimportant) that it is likely to have occurred purely by chance.

i.e. not anthropologically induced warming, but random warming (if it is a warming at all and not a random occurrence).

By Devils Advocate (not verified) on 14 Feb 2010 #permalink

By saying the warming is NOT statistically significant, it is the same as saying the warming is so small (& utterly unimportant) that it is likely to have occurred purely by chance

You repeat your earlier mistake by saying "so small" and double it with "unimportant", and compound it by "likely to have occurred purely by chance".

Question: HOW likely is it to have occurred by chance, based on what we know? And when you say "likely", what probability do you have in mind?

By Lotharsson (not verified) on 14 Feb 2010 #permalink

And a follow-up to Devil's Advocate:

If (for the sake of argument) there is statistically significant warming over some given period of time, how do you know how much is anthropogenically generated?

If, on the other hand there is no statistically significant warming, does your answer change? Does it change if there is warming that doesn't quite reach statistical significance? Does it change if there is cooling that DOES reach statistical significance? In other words, if you had a parallel universe to play with and took away the anthropogenic influences, what difference in temperature would you expect to see? Why?

By Lotharsson (not verified) on 14 Feb 2010 #permalink

Must be a touchy subjectTim is blocking my posts.

By Devils Advocate (not verified) on 14 Feb 2010 #permalink

[Bedevilled arrogant](http://scienceblogs.com/deltoid/2010/02/daily_mail_caught_in_another_l…).

Let's try explaining it to you this way⦠Where one has a consistent warming signal overlaid with noise, one requires a period of time to discern the signal from the noise.

It's simple, really, if you think about it.

If the magnitude of the noise (essentially the range of the residuals of a regression line spanning a representative period of the warming) is greater than the magnitude of the signal for a particular interval, one requires a greater amount of time than said interval to statistically detect the signal.

It does not mean that there was no underlying warming occurring during the interval. If one attempted to use you logic, one could divide an obvious warming trend into a series of smaller intervals, where the noise swamped the signal within each interval, and say that because there is no statistically significant warming in any of the intervals, there is no warming at all.

How completely bogus, and how patently ignorant of statistical trends.

You must have practised very hard, and for a very long time, to reach the level of idiocy that you display.

I suggest that you go right back to grade 1 and start your education all over again. Something might actually stick the second time around.

By Bernard J. (not verified) on 14 Feb 2010 #permalink

"Question: HOW likely is it to have occurred by chance, based on what we know? And when you say "likely", what probability do you have in mind?"

Based on that Jones is saying it is statistically insignificant, the warming is more likely to have occurred by chance than to have not have.

AGW is more likely to be bogus than not.

By Devils Advocate (not verified) on 14 Feb 2010 #permalink

[Devils [sic] advocate](http://scienceblogs.com/deltoid/2010/02/daily_mail_caught_in_another_l…).

Further to your fixation with the magnitude of 'differences', are you able to explain to the readers here why a very 'small difference' in a response variable of a time-series might be highly statisitically significant, and why a very 'large change' in the response variable of another time-series might not be statisitically significant at all?

If you cannot do this, then you cannot make [the claim that anthropogenic global warming is a fraud](http://scienceblogs.com/deltoid/2010/02/daily_mail_caught_in_another_l…).

By Bernard J. (not verified) on 14 Feb 2010 #permalink

"Based on that Jones is saying it is statistically insignificant, the warming is more likely to have occurred by chance than to have not have."

Here is what Jones said:

B - Do you agree that from 1995 to the present there has been no statistically-significant global warming

Jones: Yes, but only just. I also calculated the trend for the period 1995 to 2009. This trend (0.12C per decade) is positive, but not significant at the 95% significance level. The positive trend is quite close to the significance level. Achieving statistical significance in scientific terms is much more likely for longer periods, and much less likely for shorter periods.

So Jones is saying that it's very close to the 95% confidence level, perhaps "only" 90%, meaning there would be a 10% chance the measured warming occurred by chance. Thus, your claim is wrong.

Independent measurements from RSS and UAH reveal similar trends. GISS and NCDC surface records, which, unlike HadCrut, include the Arctic, reach statistical significance.

Based on that Jones is saying it is statistically insignificant, the warming is more likely to have occurred by chance than to have not have.

Er, no! Until you learn why you will continue to display your ignorance.

And please answer my other questions - they illustrate other fallacies in your argument.

Unless of course you are more interested in not changing your mind than in being right.

By Lotharsson (not verified) on 14 Feb 2010 #permalink

@52;
"meaning there would be a 10% chance the measured warming occurred by chance. Thus, your claim is wrong."

No, something is called statistically significant if it is unlikely to have occurred by chance. By Jones saying it is not statistically significant, he is saying it is more likely to have occurred by chance.

By Devils Advocate (not verified) on 14 Feb 2010 #permalink

Details matter:

a result is called statistically significant if it is unlikely to have occurred by chance.

(My emphasis.)

What is the probability (as a percentage) of something that is "more likely than not"?

What is the probability (as a percentage) of an unlikely occurrence (especially as defined in terms of statistical significance? What level of statistical significance was Phil Jones referring to? Does it match your assertion that he meant "more likely to have occurred by chance"?)

Some hints may be found here, including thought provoking quotes such as:

The use of the word significance in statistics is different from the standard one, which suggests that something is important or meaningful. For example, a study that included tens of thousands of participants might be able to say with very great confidence that people of one state are more intelligent than people of another state by 1/20th of an IQ point. This result would be statistically significant, but the difference is small enough to be utterly unimportant.

and

The significance level is usually denoted by the Greek symbol, α (alpha). Popular levels of significance are 5% (0.05), 1% (0.01) and 0.1% (0.001). If a test of significance gives a p-value lower than the α-level, the null hypothesis is rejected. Such results are informally referred to as 'statistically significant'. For example, if someone argues that "there's only one chance in a thousand this could have happened by coincidence," a 0.001 level of statistical significance is being implied. The lower the significance level, the stronger the evidence required. Choosing level of significance is an arbitrary task, but for many applications, a level of 5% is chosen, for no better reason than that it is conventional.[3][4]

(My emphasis.)

Care to change your mind on any of your assertions yet?

By Lotharsson (not verified) on 14 Feb 2010 #permalink

@55 DA

That'd only be true if for some warped reason you were using a 50% significance level.

In science, we like to be a little more sure about what we say, so generally we go for 95%.

[Bedevilled Arrogant](http://scienceblogs.com/deltoid/2010/02/daily_mail_caught_in_another_l…).

On the very page that you linked to is an explanation of why you are so wrong that you're not even wrong.

In your ignorant misrepresentation of statistics you are essentially making a [type II error](http://en.wikipedia.org/wiki/Statistical#Error).

Read [more](http://en.wikipedia.org/wiki/Type_I_and_type_II_errors#Type_II_error) and learn.

By Bernard J. (not verified) on 14 Feb 2010 #permalink

I see Bedeviled has [confirmed](http://scienceblogs.com/deltoid/2010/02/daily_mail_caught_in_another_l…) several times my [initial assessment](http://scienceblogs.com/deltoid/2010/02/daily_mail_caught_in_another_l…) that he dosn't understand what significant means.

>*BTW my other posts are being censored or blocked.*

Mine too Devil, I assume that there is an automatic block sprung when we assert each is practicing "fr*d". My blocked post [@44](http://scienceblogs.com/deltoid/2010/02/daily_mail_caught_in_another_l…) is now up, how about yours?

About the only thing to add is that using Jones' calculation, an actual trend of 0 in the HadCrut set is "very close" to having about a 5% chance (a little higher) of occuring, with the measured trend being 0.12 C per decade. Assuming a normal distribution, then the actual trend of 0.24 C per decade would be equally likely. There is that pending ship/buoy correction for SSTs that might bump the trend up a bit.

Of course, as mentioned before, we have other indicators - the GISS and NCDC surface products, and other entirely independent indicators like the satellite record and net glacier depletion, so the relevance of how great the confidence level of one particular data product over a time period of 14-15 years isn't much.

The BBC Q&A mentions many of the questions were posed by "skeptics". The purpose of some of the questions (this on in particular) seems to be to provide talking points for the denial crowd. Jones provided good technically-sound answers to these questions that are in no way inconsistent with the established science. He strikes me as being not particularly savvy with media relations, as some of the answers could and should have been expanded on. I wouldn't have bothered with the "statistically significant" question, which wasn't very relevant, unless I was ready to provide a detailed explanation of what the term means and exactly what it implies. One needs to anticipate how political hacks might distort your words. That said, I think it's also fair to say deniers are ready to distort, misrepresent, and spin nearly anything a scientist says. I hope this doesn't lead scientists to further steer clear of media relations. I believe there was a science communications presentation at the recent AGU conference. I think someone in Jones' position could work on such skills.

Lotharsson.

What ever your definition, "choosing a level of significance is an arbitrary task, but for many applications, a level of 5% is chosen, for no better reason than that it is conventional". Tests of statistical significance are harmful to the development of scientific knowledge; as per your link in pitfalls

"even when done properly, statistical significance tests are of NO value. A number of attempts failed to find empirical evidence supporting the use of significance tests. Tests of statistical significance are harmful to the development of scientific knowledge because they distract researchers from the use of proper methods Armstrong suggests authors should avoid tests of statistical significance; instead, they should report on effect sizes, confidence intervals,replications/extensions, and meta-analyses."

Maybe, if Jones could find his lost data and release it, it could be properly peer reviewed. To say the science is settled is a joke.

By Devils Advocate (not verified) on 14 Feb 2010 #permalink

To take a step back, it's kind of astounding that the reason Jones is getting attacked on this is that he expressed a perfectly correct level of doubt about just how certain a particular claim about climate change might be. Indeed, his point is that even though the numbers clearly point towards a particular conclusion, as a scientist he won't say that this conclusion is certain without more evidence.

And for this, he's being told he's basically a cultist.

I only got a B minus and it wrecked my GPA, but damn am I glad in retrospect I had to take a sophomore level statistics course for my BA. The basic concepts in this "debate" are really not all that hard to grasp, unless you're either totally stupid or deadly determined to delude yourself or others.

Based on that Jones is saying it is statistically insignificant, the warming is more likely to have occurred by chance than to have not have.

This is golden. Just for the record, Devils Advocate, do you actually, really, truly believe this?

Where did you study statistics (if at all)?

By Martin Vermeer (not verified) on 14 Feb 2010 #permalink

DA quoting:

"even when done properly, statistical significance tests are of NO value. A number of attempts failed to find empirical evidence supporting the use of significance tests. Tests of statistical significance are harmful to the development of scientific knowledge because they distract researchers from the use of proper methods Armstrong suggests authors should avoid tests of statistical significance; instead, they should report on effect sizes, confidence intervals,replications/extensions, and meta-analyses."

Just below this was a quote that DA failed to give:

"Use of the statistical significance test has been called seriously flawed and unscientific by authors Deirdre McCloskey and Stephen Ziliak. They point out that "insignificance" does not mean unimportant, and propose that the scientific community should abandon usage of the test altogether, as it can cause false hypotheses to be accepted and true hypotheses to be rejected."

In the present case "insignificant" is the trend from 1995 to 2009 and as McCloskey and Ziliak point out, that does not mean unimportant. As they also point out, it can cause a true hypothesis, such as global warming, to be rejected. In this case by people who have a political axe to grind.

However, if significance means something than it's easy to achieve it by just extending the sampling period to 20 or more years.

By Chris O'Neill (not verified) on 14 Feb 2010 #permalink

Devil's advocate was for statistical significance before he was against it ;-) Priceless!

I do wonder what alternative (s)he would propose - given that most of modern medicine relies on it. Or does (s)he eschew any treatment developed after the 1800's?

...instead, they should report on effect sizes, confidence intervals,replications/extensions, and meta-analyses

Yeah, kind of like the IPCC did in their reports.

I look forward to Devils Advocate rejecting confidence intervals and meta-analyses and the like too ;-)

By Lotharsson (not verified) on 14 Feb 2010 #permalink

It's interesting that the denialists suddenly believe the prefix "Phil Jones claims..." is the mark of TEH TRUTH.

It's /almost/ as though they'll listen to anyone who says what they want to hear.

Neil, they have a diode filtering all input to the brain. It only admits information that supports what they already think ;-)

By Lotharsson (not verified) on 15 Feb 2010 #permalink

As was pointed out above, HadCRUT is not the only global temperature record. I did a calculation for the years 1995-2009 using GISTEMP annual means and got a warming of 0.15°/decade with a t of 2.96 and 13 degrees of freedom, which is significant at p<0.025. This episode is almost a litmus test; people who know a little bit about stats and trend analysis will find Jones' statement unremarkable. The ignorant will turn this into a big deal.

Deech, you're dead on with your last statement. This "news" also made the front page of today's Daily Express here in the UK. They said something along the lines of 'Climategate scientist admits there's been no global warming for the last 15 years', which as we know is not strictly (read: remotely) what he said.

Most would consider the Express an even worse rag than the Mail. As comedian Russell Howard characterised their style: "Don't go outside! It's full of queers, blacks and crime! Oh, if only Diana was here."

BTW, the Express is famous for trumpeting the "forecasts" of Piers Corbyn, who is such good fun it's almost sad to see his predictions consistently failing (unless you're Piers himself, in which every forecast of a devestating storm is verified by a 70mph gust on a remote Scottish Island).

Perhaps Tim could do a post or two on Piers, given that he's an ardent AGW sceptic who gets a reasonable amount of attention here in the UK? If nothing else, it'll give us all a laugh.

Stu, Corbyn was atthe Royal INstitute discussion on extreme weather and AGW a couple of weeks ago. He made sure to ask a big rambling question which made little sense, interrupt with comments a few times, and claim an 85% accuracy rate, not to mention promoting his services at every opportunity.

When I saw the title of this post I had to ask myself where the news was. The Daily Mail being caught in a lie is like the sun being caught rising in the morning. The credibility of the Daily Mail has taken another body blow? How? How do you land a blow on something that doesn't exist?

Now if the Daily Mail had been caught printing well researched and reality based journalism, now *that* would be news.

By Captain Obvious (not verified) on 15 Feb 2010 #permalink

You are all looking at trees and missing the forest. CO2 levels continue to rise rapidly and yet over the last 8+ years the planet has cooled slightly. Unless I missed something, your theory relied on warming happening due to CO2 levels rising ... remember Al Gore showing us all that graph where the temperature rising when the CO2 level did? Add to that the raw data which is available to everyone (the BOM site here in Australia ... check out their Australian average before looking at the individual stations and try and find more then a few that match their graph let alone go high enough to justify the average) showing nowhere near the same level of warming. Why not do it just to prove me wrong.

You are all looking at trees and missing the forest. CO2 levels continue to rise rapidly and yet over the last 8+ years the planet has cooled slightly.

I know, it's so disappointing. All other forcings and feedbacks were simply standing aside and letting CO2 have its way with the world and here it turns out it's impotent, eh? Well, there's politeness for you.

Okay sun, ENSO and whatever else. It's time to start having an effect again! You needn't wait for CO2.

I'm just trying to figure out what changed between 1994 and 1995 that justifies excluding 1994 (and earlier) from the trend analysis. Why stop at 1995 when we have measurements from 1994 too?

PS #81, don't ask silly questions ;-)

By Martin Vermeer (not verified) on 15 Feb 2010 #permalink

Why stop at 1995 when we have measurements from 1994 too?

I've pondered that, and I think I know why.

Because if you're a cynically manipulative skeptic, you expect to be able to take advantage of the scientist's tendency to precision in communication in order to get them to say something that you can spin to imply they said something else. So you calculate that 1995-2009 falls just under statistical significance, which is a term that you know most people don't understand, let alone understand the ... er, significance of, so it gives you the room to spin that you need because the facts aren't on your side.

You could have asked if there's been any warming in the last 15 years (1994-2009), given that (IIRC) Tamino shows that 15 years is about the minimum for reliably detecting a warming signal if conditions are on your side (and 30 years are much better).

So you ask about the last 14 years instead.

By Lotharsson (not verified) on 15 Feb 2010 #permalink

Unless I missed something

You're missing a lot, Joe. Tell you what, you go and find a passage in any IPCC document which says that CO2 is the only factor affecting temperature change, and I'll do the rest of your homework for you.

Could someone please clarify one point for me, as I have no qualifications whatsoever in statistics. Given that the 0.12C warming trend identified by Jones falls (just) outside statistical significance, does that mean that any uncertainty is equally possible on the up side as on the down side?

>*Given that the 0.12C warming trend identified by Jones falls (just) outside statistical significance, does that mean that any uncertainty is equally possible on the up side as on the down side?*

Absolutely, as MarkB [points out](http://scienceblogs.com/deltoid/2010/02/daily_mail_caught_in_another_l…) and as is [shown by Tamino](http://tamino.files.wordpress.com/2009/12/gissrat2.jpg) the chance of zero warming in the last 14 years has the same probability as the last 14 years having warming trend of just under 0.3 degrees/decade.

Andrew - I am a neophyte when it comes to stats, but what it means is that the trend does not meet the criteria of being less than 1 in 20 likely to be random. At the moment, since it does not meet the 95% criteria, there is a 1 in 15 or 1 in 10 or whateverchance that that short trend came about by accident. (Of course this says nothing about the various short term drivers of temperature, and the stats cannot say anything)
So the uncertainty is not about the sign of the number, but about whether it came about by chance or not.

83 Lotharsson,

I think you are correct in your reasoning, in that the period was carefully chosen to produce a resply which would be misrepresented, but I'm not sure about the arithmetic. Are those years inclusive, i.e., Jan of the start year to Dec of the finish year? If so, 1995-2009 *is* 15 years, and 1994-2009 is 16; if not, they are 14 and 15, respectively, as you say.

This is something to watch out for at WoodForTrees, where the From year is included but the To year is not.

By TrueSceptic (not verified) on 16 Feb 2010 #permalink

TrueSkeptic, yes it depends whether you want to count intervals or years of data, and in the latter case whether the full year is included.

But I reckon no matter how you choose to count it they specifically chose one too few years to achieve statistical significance.

By Lotharsson (not verified) on 16 Feb 2010 #permalink

Good grief. As an example of the echoes the Daily Fail caused in the padded cell world of 'blog science', just check out this quick exchange I had with SBVOR, who hangs around the accuweather global warming blog, suckling at the teat of their scepticism (to be fair, accuweather may seem to have a sceptical stance, but are far, far more balanced that WUWT, CA etc etc...)

Check it, reproduced in full (everything in italics):

Stu said...

But SBVOR, NOAA said:

"The [computer model] simulations rule out (at the 95% confidence level) zero trends [in global temperatures] for intervals of 15yr or more"

And Jones said:

"I also calculated the trend for the period 1995 to 2009. This trend (0.12C per decade) is positive, but not significant at the 95% confidence level".

Can you see the difference? NOAA say that if the observed trend is zero (nada, nothing, flat, not 0.12C/decade) for 15 years, then you can reject AGW as currently formulated at the 95% confidence level.

Since the observed trend is actually 0.12C/decade.. well, you can't reject the hypothesis with 95% confidence.

Do you retract your, arrogant, plain wrong, emboldened and overly capitalised conclusion?

Feb 16, 2010 2:27:00 PM
SBVOR said...

Stu,

Nope, no retraction required or forthcoming.

By Philâs own admission, the observed warming is NOT âstatistically significantâ. In other words, the observed warming (from datasets PROVEN to contain âan estimated warm bias of about 30%â) falls within the margin of error and cannot be said -- with any confidence -- to exist.

Ergo, by NOAAâs own proclamation, the IPCC models have been invalidated.

End of story -- time for you to find a new cult!

I mean, what can you when trying to argue with someone who a) knows nothing about statistics and b) refuses to learn anything about statistics? (S)he's as bad as Devil's Advocate.

from datasets PROVEN to contain âan estimated warm bias of about 30%â

Eh? Where's that figure/quote pulled from, assuming it's not SBVOR's arse?

Bud, I think you're on to something! I reckon he polled his arse for current temperature (approx 37 degrees C), assumed the data sets came from similarly situated sources, and compared 37 to the daytime temperature on the thermometer in his back yard which was probably about 28.

(37-28)/28 - yep, about 30%.

Q.E.D.!

By Lotharsson (not verified) on 16 Feb 2010 #permalink

Bud, apparently it's something Pielke said:

Enjoy the rest of teh crazy on that blog by the way, it's good fun!

I'm in a debate about this with a guy who knows way more about statistics than I do (knowing nothing). I've been reading up a bit, but have a question. I warn you, could be a stupid one!

On what is the p-value based when calculating statistical significance for a certain period of observation of a certain dataset, in this case the 1995-2009 period for HadCRUT?

In other words: why is 0.15 deg. C per decade deemed statistically significant and 0.12 deg. C per decade not?

Neven:

In other words: why is 0.15 deg. C per decade deemed statistically significant and 0.12 deg. C per decade not?

According to , statistical significance can be considered to be the confidence one has in a given result where the confidence is given by:

confidence = (signal/noise) â(sample size)

So the greater the signal, the greater the confidence. The signal 0.15°C is greater than 0.12°C, so the confidence is greater simply because it's a larger signal.

By Chris O'Neill (not verified) on 18 Feb 2010 #permalink

@Neven: If the confidence interval is, say, ±0.14 C/decade, then a trend of 0.15±0.14 C/decade is a significant upward trend, whereas a 0.12±0.14 C/decade is not. Basically, there's non-negligible odds that it is not an upward trend.

People, we need to understand what is behind this affair. The BBC asks Prof. Jones a number of questions in writing "including several gathered from climate sceptics". One of the question is were craftily formulated, and the one asking it must have been fully aware of what the only possible (true) answer was. And by pure coincidence, the Daily Mail can reinterpret that answer into a simple talking point everyone can understand. Of course it would have been even better if Jones hadn't answered the way he did, and they could have opened with "Prof. Jones caught lying again." Now how statistical significant is this chain of events?

By Lars Träger (not verified) on 18 Feb 2010 #permalink

Thanks for the answers!

Joseph wrote: "If the confidence interval is, say"

That's the point, I believe. What IS the confidence interval used to determine statistical significance? Is it based on something like an average of the baseline period?

The guy I'm discussing this with is asserting that the null hypothesis for the period of 1994-2009 is that the warming trend will be less or equal to 0.12 deg. C/decade. The alternative hypothesis is that the trend will be greater than 0.15 deg. C/decade. There's a flaw in his thinking, but I'm really zero at statistics so I don't know what it is.

The guy I'm discussing this with is asserting that the null hypothesis for the period of 1994-2009 is that the warming trend will be less or equal to 0.12 deg. C/decade.

Uh, I do believe that the null hypothesis is that the observed data is simply due to natural variation ... in other words, you can't reject to the 95% confidence level the possibility that there's no warming trend at all during that time frame, and what's observed is just the noise in the system.

Of course, Jones hints that you can reject it to a confidence level *just shy* of 95%.

You could ask your statistics "expert", if, given two bottles, he was told that one contains 95 capsules of cyanide and 5 of powdered milk, and the other 93 capsules of cyanide and 7 of powdered milk, and he was given three choices:

1. chose a capsule from the first bottle and swallow it.

2. chose a capsule from the second bottle and swallow it.

3. buy a cheeseburger from McDonald's.

Which choice would he take?

Dhog,

rather than believing that the null hypothesis is that there's no real trend, I can state the following:

The null hypothesis should always be that there's no actual trend and that the observed pattern is an artefact of the noise. The less noise, more trend and more data you have, the more confidence you have that the trend is or isn't by chance.

Neven says: "What IS the confidence interval used to determine statistical significance? Is it based on something like an average of the baseline period?"

The confidence interval is always 95%, which is very close to two standard deviations. Plot two standard deviations above and below your mean and see if the observed trend goes outside this range. If it does, you have statistical significance at 95%.

Thanks for all the explanations. This guy is assuming a lot of things without looking into them (for instance, he constantly talks about the IPCC whereas it's CRU's dataset we're talking about) and this leads him to conclude that figures are made up out of thin air. Unfortunately I lack specific statistical knowledge to explain to him what the flaws in his thinking are, so I won't waste my time going into it. It's hard enough to convince him about things I do know something about.

@Nevin,

It's kind of sloppy thinking of it this way, but in the 'null' language you're using:

The 'null hypothesis is that the true trend is zero. The observed trend is 0.12 C / decade. Jones said that we aren't quite 95% sure that we can rule out the possibility that the observed trend for the last 15 years is an artifact of noise in the system.

However, the confidence interval is on both sides of the observed trend - we also aren't quite 95% sure that we can rule out an actual trend as high as 0.24C / century, with noise making it look as low as 0.12C.

Jones is a very good scientist - but he's piss poor when communicating with those who want to twist his words to make him look bad by misrepresenting him. I find it hard to fault him for that.

Jones is a very good scientist - but he's piss poor when communicating with those who want to twist his words to make him look bad by misrepresenting him. I find it hard to fault him for that

I'll fault him for offering to submit to the Q&A in the first place. It wasn't a phone interview, he was given the answers and answered them in writing, apparently. He could've gotten help from someone more astute (in regard to the press), at least.

As I said on the thread over at island of doubt, it's a good thing to point out the symmetry, that the less than significant observed +0.12C/decade trend over 15 years makes a trend of +0.24C/decade as likely as 0C/decade ...

Keep hammering that point home.

Re dhogaza @ 100;

[The paradoxical effect can be seen because even though the test is "99% accurate", if one receives a positive response from the test there is still a 99% chance that the test gave an incorrect result."](http://en.wikipedia.org/wiki/False_positive_paradox).

By Devils Advocate (not verified) on 21 Feb 2010 #permalink

That would be Devil's - [statistically significant](http://scienceblogs.com/deltoid/2010/02/daily_mail_caught_in_another_l…) means [50% probability](http://scienceblogs.com/deltoid/2010/02/daily_mail_caught_in_another_l…) - Advocate;

And Devil's - I was for statistics until they [didn't say what I wanted](http://scienceblogs.com/deltoid/2010/02/daily_mail_caught_in_another_l…) - Advocate;

And Devil's - Louis Hissink is an [intellectual giant](http://scienceblogs.com/deltoid/2009/12/plimer_exposed_as_a_fraud.php#c…) - Advocate.

Ever considered changing your name to Dunce's Advocate?

Reference is often made to the 2500 unanimous scientists who agree on the conclusions in the IPCC Report of 2007. However, during a meeting in Akersberga, Sweden Mr Bert Bolin pointed out that only 400 of them were climate specialists.

Bearing in mind that there are more than 9000 scientists, many of them climate specialists, who have stated that they are sceptical to the conclusions in the IPCC Report, it would be very interesting to know who those 400 supporting the conclusions are, or were, not to mention the majority of all the climate scientists who are said to support the conclusions in the IPCC Report

@Yngve, I'd say, start listing the people you believe are climate "specialists". Be aware that people like "Ian Plimer" or "Hans Labohm", or "Arthur Rörsch" or "Sonia Boehmer-Christiansen", or "Benny Peiser" or "Lubos Motl", or "Jaworowski" (I always forget the guy's first name) and a long list of others is going to result in eternal ridicule.