Bishop Hill Not answering the question
Apr 17, 2013 Climate: MetOffice Climate: Parliament Graham Stringer's question to the department of Business Innovation and Skills has received a response, but not an answer. This from Hansard:
Graham Stringer: To ask the Secretary of State for Business, Innovation and Skills whether the claim that (a) every year since 1998 has been significantly warmer than the temperatures you would expect if there was no warming and (b) for the last three decades the rate of temperature increase is significant made by the Met Office in a climate science briefing sent to the chief scientific adviser on 8 February 2010 was supported by any statistical time-series analysis. [150533]
Michael Fallon: The full statements sent by the Met Office to the chief scientific adviser on 8 February 2010 are (a) every year since 1998 has been significantly warmer than the temperatures you would expect if there was no warming (baseline of 1861-1900) and (b) for the last three decades, the rate of temperature increase is significant even when uncertainties in the observations are factored in.
These statements are based on analysis of HadCRUT3, the global temperature dataset compiled by the Met Office and the university of East Anglia’s climatic research unit.
Reading between the lines I think we can probably say that the advice Julia Slingo has been providing to central government is not based on time series analysis.
Oh dear.
Reader Comments (33)
The question clearly and specifically asked whether they were supported by a time-series analysis.
Can the minister get away with simply ignoring the question?
Must Graham Stringer do a Paxman?
Wikipedia on the US 5th Amendment:
Fair enough, but I don't think it is supposed to apply to questions asked of ministers in Parliament.
The minister's behavior is reminiscent of some mechanically ignorant car salesman who's memorized every line in the sales literature without the slightest clue what they actually mean.
It isn't, of course, supported by a time series analysis. I had a brief exchange with Doug McNeall on his blog about this issue. He told us, on the blog, that the increase between 1880 and 2000 was "scientifically significant*, I tried to understand what this actually meant by asking Doug if he'd applied the same parameters of scientific significance to the period 1880 -1940 would that be "scientifically significant". He described my questions as "interesting" but didn't answer them. I'm left to assume that there are no parameters to scientific significance, it's just your normal climate science "gut feel" masquerading as science.
He did seem to get into a bit of a huff over my assertion that GATA wasn't really a useful parameter unless we are comparing the current anomolies with the average measured of the base line 1961n- 1900 unless both sets of measurements were made on precisely the same thermometers normalised for any changes in the period since. Told me I could help out if I wanted to, but later said he'd not taken, nor intended to give offence.
Must be great in the Met Office canteen.
Something tells me that the Met Office canteen discussions focus around retirement resorts and conferences in warm places - plenty of options for people with large travel budgets and 170,000 pounds year - a pretty fair rate of pay for ignoring questions containing the word 'significant'. Just have to hope that Heathrow isn't snowed in when you make your getaway, right Julia? (No point in reading the forecast, of course).
Dodgy Geezer
"The question clearly .... asked ...."
Well, actually, no it didn't. I know what the question said; you know what the question said.
I just ran it past Mrs J — who, I can assure you, is every bit as bright as I am — and it needed three goes before she got the point.
This is not to excuse the Met Office but it does excuse the Minister. It is a complex question (I speak as someone who earned his living from the English language) and the thrust of it only becomes apparent at the very end.
One could easily be forgiven for thinking that the Minister was being asked if the Met Office had made the claims that were being stated rather than whether they had used time-series analysis to reach their conclusions.
In fact the first time I saw the question I also needed two goes before it hit properly.
It is just the sort of question that gives civil servants every excuse to obfuscate — whether deliberately or otherwise — and just because we grasped the meaning (we did?) is not proof positive that everyone else did.
4/10, Graham. Re-write and re-submit:
"To ask the minister whether the Met Office employed a statistical time-series analysis to reach the conclusions in their briefing in 2010 which stated that ......
... and if not, why not?"
See the difference?
PS
If this was a written answer then the chances of it having passed in front of the minister's eye are slim; through his brain, even less so.
Mike Jackson - a valid point. If left as Stringer structured it, at the very least it could use a couple commas.
Mike Jackson
I agree the wording of the question is almost impenetrable. It is always a good idea to assume that in politics the respondent will use every excuse to answer a different question from the one you asked so don't give him/her any excuses.
4/10 is over-generous.
By the by, you are not the Mike Jackson of BD&F fame?
Having once played a very small part in drafting replies to PQs, I would have thought that the import of the question is whether the claims etc were supported by any time series analysis and that is clear enough and would have had to be included in the draft answer. Once we minor bods had drafted an answer, it would go to the Minister to make the reply, even a written reply. So I would expect him to have read it, even cursorily.
Does Slingo know what day it is?
Bish - you have friends at the Met Office. Some read this blog. Perhaps they can intervene and clear out this unfortunate misunderstanding.
If you want to know if two years are significantly different in the Hadcrut4 data look at the 95% confidence bounds of the smoothed blue curve for the Combined land and Sea surface graph. If the 95% confidence limits overlap, the difference is not significant . If they do not overlap, the two dates are significant.
http://www.metoffice.gov.uk/research/monitoring/climate/surface-temperature
For the two dates you mention the difference is not significant. Please complete your argument, and explain why you chose those particuar dates.
Bouncing around a bit. That last comment of mine was intended for geronimo on "And another".
Can you lot not read? The Minister's answer boils down to a) yes b) yes.
EM. I don't think you've quite grasped the meaning of statistical significane. Have another go, but don't ask at the Met Office canteen, I don't believe you'll find the answer there.
EM - "you lot" is rude. Please be more polite.
I know that schoolteachers often address their classes as "you lot" but this does not make it any less rude.
@Mike Fowle
...Having once played a very small part in drafting replies to PQs, I would have thought that the import of the question is whether the claims etc were supported by any time series analysis and that is clear enough and would have had to be included in the draft answer....
Indeed. I also have answered PQs, and it is clear that the purport of the question involves time-series. It would have been clear to the people who drafted the initial response. But in true 'Yes, Minister' style it is possible to interpret the question as being only about whether the stats were based on 'an analysis of some kind' - and so that is how it was presented at the higher levels.
This will not have been missed in the house - they are on the back foot. Of course they will always provide an answer somehow - simply cast the decision about the appropriate statistical technique to use as an internal technical one, and say that the Met Office used their teams of statisticians to apply numerous analyses, some of which would have been time-series, and that the final advice which came out was the sum of their best technical judgement.
Stringer's questions were very poorly framed. He is lucky to have got any answers - let alone misleadingly, vacuous ones.
I rather agree with Bernie. Stringer could have been more direct.
But I think we are seeing the vast edifice of Whitehall climate-change certainty, one in which you will never get ahead if you question even slightly the consensus, beginning to creak.
Given the vast numbers of the smug, reared on the absolute belief that Global Warming is real, is here, that the science is settled and that huge numbers of centrally funded jobs depend on it, a long haul can be guaranteed.
So we are still a long way from an utter rout, which will I am sure be the final result. But the damn is beginning to look distinctly vulnerable. Wait for the moment when, suddenly, it collapses absolutely, sweeping its proponents into sudden, resentful oblivion.
At which point, we should all do no more than politely point out that they were wrong at the start and that their final disintegration should have been no more than the expected consequences of advocates suborning science in the interests of political goals.
"Time series analysis" just means some sort of analysis of a time series (i.e. a list of data points each associated with a sequential time). *Any* analysis of historical weather data is a "time series analysis". You need to be specific about what sort of analysis you intend.
What was meant was that there is a standard body of statistical methodology for analysing time series commonly known as "time series analysis", and the question was about whether these standard methods had been used. But as written, it's like asking whether "weather data analysis" was used. Of course it was - even if all you did was look at the weather at the start and end of the period and subtract.
You need something like:
"The Met Office claimed in a climate science briefing sent to the chief scientific adviser on 8 February 2010 that (a) every year since 1998 has been significantly warmer than the temperatures you would expect if there was no warming and (b) for the last three decades the rate of temperature increase is significant. We ask the Secretary of State for Business, Innovation and Skills what statistical model was used to determine the distribution of temperatures to be expected without increased anthropogenic CO2, what statistical model was used to declare the rate of temperature increase "significant", and to provide public access to the formal verification and validation of these models."
ntropy wins
That would seem to be a correct assumption since I don't understand the reference.The French have only ever heard of one Michael Jackson which has had its amusing moments especially when going through Passport Control — amusing for the Customs Officials that is.
geronimo
At the level discussed here I use this rule of thumb.
Two sample means may differ by chance, or because something has caused them to be different.
If the probability of a difference by chance is less than 5%, one may reasonably infer a cause for the difference.
95% confidence limits are a statement that the probability of the actual mean of the population sampled, as opposed to the sample mean, being outside that range is less than 5%.
If the 95% confidence limits for two sample means do not overlap, the probability that the difference between the means is due to chance is less than 5% and therefore statistically significant.
The 95% confidence boundaries for the smoothed curve means for 1880 and 1940 overlapped, so the probability that the difference between the means was due to chance exceeded 5% and the difference between them was not statistically significant.
There are more sophisticated measures of significance, but they require access to the full dataset and a lot more time and effort to carry through.
What methods have you used to analyse the Hadcrut data and what statistical significance have you calculated for the difference between 1880 and 1940 global mean temperatures?
Martin A
What is a suitable collective noun for those of your persuasion? Having been called many things from arrogant to ignorant to insane by the more unpleasant denizens of this blog I am becoming much less inclined to politeness and much more inclined to give measure for measure.
My apologies if "you lot" offended you personally, as you argue using science rather than by insult. Consider it the response of an exasperated teacher to a class perfoming well below its potential.
"The 95% confidence boundaries for the smoothed curve means for 1880 and 1940 overlapped, so the probability that the difference between the means was due to chance exceeded 5% and the difference between them was not statistically significant."
That's not a safe thing to do.
Take two normally distributed random variables X ~ N(-1.9,1) and Y ~ N(1.9,1). (That means Normal distributions with standard deviation = 1 and means -1.9 and 1.9 respectively. The 95% intervals (+/-1.96 either side of the mean) overlap.
However, the difference (Y - X) is distributed N(3.8, 1.414), which has a probability of 0.0036 of being equal to or less than zero. Even though the 95% intervals overlap, they are significantly different. You can think of it as being because to be near the tail of one distribution is unlikely, but to be near the tail in *both* distributions is unlikely squared: which is *very* unlikely.
I realise you said you needed to do a more sophisticated analysis, so I expect you knew that, but you need to be careful about how you word it when non-mathematicians are listening. :-)
Hansard is full of incisive questions and abstruse answers. It's a bit of a theatrical charade.
EM
Or you could of course just go away and leave us in peace.Like your fellow trolls Zed and BitBucket you seem to think that you have some God-given obligation to "correct" what you see as our "misconceptions".
I will grant that your contributions do usually make some attempt to address the question but the de haut en bas superciliousness and general attitude of faux superiority does tend to piss people off.
As far as Stringer's question is concerned (which I think is the subject of this thread) the answer did not address the question which was "did the Met Office use time series analysis to come up with 2010 briefing figures?", not "was the briefing accurate in these two specific respects" which is the question that your "yes" and "yes" is answering.
Thank you for the compliment.
Personally I was not offended - I have been called all sorts of things over the years but I learned to use it as useful input about the person making the remark, rather than taking it personally.
As I rather imagined, it seems you don't see the phrase as especially rude and you see it as permissible to use when exasperated. But, nonetheless, it is a disparaging way to address a group of people and therefore intrinsically rude.
I am sorry that some BH commenters have used offensive terms in addressing you. But, because of the nature or a blog, that is down to them individually, rather than to all who comment here.
To me "you lot" does not sound that much different from "you *****s" (insert whatever unpleasant word you wish).
I made the comment simply because rudeness inhibits profitable discussion and should (I believe) be avoided on BH in the interest everyone gaining the maximum benefit from the exchanges.
Added after initial posting:
"What is a suitable collective noun for those of your persuasion?"
I think it's misconception that all BH posters hold a common set of views or "persuasion". I feel sure you can make the points you want to make without needing to have a "collective noun" for people who read and comment on BH.
Mike Jackson
apologies for the rather cryptic reference
as far as I am concerned all the many Michael Jacksons in this world are fine upstanding members of society even those who have chosen to forsake their fatherland (or is it motherland?) for pastures greener
I was once almost fluent in French but then I sobered up.
Jim Hacker: "Well, you answered all my questions"
Humphrey: "I'm glad you thought so, minister".
The Establishment in the Anglo-centric world is noted for carefully-constructed dialogue that is designed to be not quite that which it appears to be. Some of it can be very funny when captured in theatrical comedy and novels but it's not so funny when it is employed to deceive or to dissemble, as it is this example. The question is clear in it's intent; the answer is slippery and evasive at best sand arguing about this has similarities to debating the number of angels that can be accommodated on the head of a pin.
Having also been involved in both framing and answering PQs in a Westminster system (although a slightly different one) I agree that the clearer and briefer the question, the harder it is to avoid answering it. The initial question was poorly drafted, IMO.
There is scope for Stringer to prepare a follow-up question which forces them to answer - such as asking for the specifics of the analysis technique used by the MO. The language should be as plain, specific and non-technical as possible.
The answer to the pinhead and dancing angels question is: an infinite number. However, since they would be bodiless spirits their dancing would not be much to look at.
My own judgment is that there can be no mindfulness (aka 'soul') without some material form and that there are neither immaterial angels nor material ones.
I have a fundamental question.
EVEN if it has got warmer over that timescale: SO WHAT..?
Met Office Annual Central England Temperature data shows:
since the Little Ice Age 1850 -1999 +0.8 degC
Since the turn of the millennium 2000 - 2012 -1.18degC.
So the last 12 years have obliterated all the warming recovering from the Little Ice Age.
However looking at winter temperatures December - March the decline is even more pronounced at -1.4degC.
Seeing as how these are Met Office data the rapid decline even for the limited coverage CET data set should be incontrovertible.