John Quiggin has just put up another post on the Phil Jones statistical significance issue. This time putting the boot into Richard Lindzen. It seems Lindzen is the originator of the claim that there has been no statistically significant warming since 1995. Jones agreed with that statement at the 95 percent level, but only just. Recall my discussion on that point here and here.
This is what John Quiggin says (emphasis added)
Lindzen has published a couple of hundred papers in climatology, so I think we can assume he knows that the statement “there has been no statistically significant warming since 1995″ means nothing more than “given the variability in the data, we need at least 15 observations to reject the null hypothesis at 95 per cent confidence”, a fact so trite as not to be worth mentioning.
It is sad to see a respected scientist reduced to this kind of thing. And as far as I can tell, all this is simply to avoid admitting that he backed the wrong horse back in 1990, when he bet that he was smarter than the majority of climate scientists who thought humans were (probably) causing global warming. The data since then has supported the majority view, but instead of revising his position, Lindzen has resorted to dishonest statistical trickery.
The problem with that argument is the bit about 15 observations – John is exactly correct on that point; using 15 observations would be a dumb thing to do. When Lindzen wrote that email he would have had less than 15 annual observations. But I suspect Lindzen and Jones used monthly data for their analysis not annual data. Using monthly data you end up with 15 x 12 = 180 data points.
At a general level the question is whether using 15 years of monthly data is enough? My gut feeling is to say no. If we believe that climate change occurs over long periods of time, we should use longer time series. That is what Terry Mills does here.
Sinc:
If we apply Mill’s argument he suggests there is nothing unprecedented about this spike in global temps as the last chart and the conclusion suggests.
Isn’t that right? Right in the sense that’s what Mills is alluding to.
JC
3 Mar 10 at 5:02 pm
I think that’s pretty much it, JC. And this is really bad news for climate scientists since the money being funneled to their research programs, which is orders of magnitude greater than the funding received by equivalent disciplines, are predicated on the current warming being exceptional.
dover_beach
3 Mar 10 at 5:14 pm
Two problems with your analysis
1. With monthly data, the standard deviation rises so the number of observations needed for significance goes up. In general, if you need 15+ years of annual data to reject the null of no trend, you need 180+ months. So, it makes no difference whether you use monthly or annual data.
2. The glaring problem is that there is no reason to cut off the data at 1995 *except* that this as far back as you can go without getting a statistically significant trend. The trend since, say, 1988, is clearly significant. That’s a good starting point since people like Hansen were already predicting climate change, so there is no risk of retrodiction or cherrypicking, as with claims like “global warming ended in 1998″.
John Quiggin
3 Mar 10 at 6:00 pm
so let me get this straight. If you analyse 1994 to the present, the warming is significant? This is a genuine question.
daddy dave
3 Mar 10 at 6:21 pm
Second question.
There is obviously a great interest – and a need – for analyses to look at increases or decreases in the warming trend. is it accelerating? Is it flattening out? and so on.
But if the kind of analysis described here is invalid, then what? What sort of methodology should we be using for detecting trend analysis? Another genuine question. There are methods around for doing this, but I wonder what JQ’s, sinclair’s et al thoughts on it are.
daddy dave
3 Mar 10 at 6:24 pm
Don’t know about. Finance types are always trying to increase the frequency of the data. But I think what does make a difference is if you’re criticising the guy for having 15 observations when he might have 180.
In principle I agree, similarly no reason to start at 1988 either.
Sinclair Davidson
3 Mar 10 at 6:24 pm
Sinclair, not sure how much “gut feeling” is statistically significant
rog
3 Mar 10 at 6:33 pm
Daddy Dave, the answer is “Yes”. If you go back to 1994 or earlier the trend is significant, a point conceded by Lubos Motl here
http://deepclimate.org/2010/03/02/round-and-round-we-go-with-lindzen-motl-and-jones/
And it’s highly likely that once the data for 2010 is in, the trend since 1995 will be significant. It’s all a matter of having a large enough sample size.
John Quiggin
3 Mar 10 at 7:37 pm
The Mills paper suffers from the same problem as your example of using monthly data. The time series for one location (Central England) has much higher variance than for the world as a whole (check the Y-axes on his graphs) so its much more likely that you will observe big movements over a small number of years, just by chance, making it harder to reject the null hypothesis. Even so, I note that he doesn’t undertake a statistical test for a break post 1970.
John Quiggin
3 Mar 10 at 7:40 pm
Finance types like high-frequency data because they are interested in high-frequency components of volatility. If they had to argue with people who claimed that the stock market hadn’t risen over time, annual data would be fine.
John Quiggin
3 Mar 10 at 8:10 pm
I’m not talking about the market microstructure stuff. I’m thinking about estimating asset pricing models with, say, monthly and weekly data.
Sinclair Davidson
3 Mar 10 at 8:33 pm
John – while monthly data has a higher standard deviation than annual data, I don’t think you can assert that it requires 180 months of data to be equivalent (in this example) to 15 years of data. It really depends on how much higher is the standard deviation. I doubt that as many as 180 months would be required to reject the null hypothesis at 95 per cent confidence.
Samuel J
3 Mar 10 at 8:37 pm
DD – if you estimate the models I did in the Oomph post the post 95 trend becomes statistically significant at the 95 percent level if you take the sample back to September 1994. But you’ll recall that PCE said to use a MA(1) term as well. When I do that the required sample starts in July 1992. The important thing is going forward – what happens when we add the 2010 data? We’ll know that in aout 11 months time. (Bear in mind that this assumes that what I’m doing is a reasonable proxy for what Jones did).
Sinclair Davidson
3 Mar 10 at 8:39 pm
Surely it makes more sense to start at a time point when the amount of CO2 humans started belching out into the atmosphere began to increase dramatically (as can be seen in the linked paper) Given this is far before 1994, I’m not sure what the point of arguing over statistical significance really is, apart from to refute silly people that cherry pick points (I assume if we took, say, 1900 as a starting point, the p value would be < .0001 or whatever it really is, and if we look for points around that date, there is really nothing to cherry pick).
.
It seems to me the only interesting thing about 1995 (or 1994) or whenever would be if you had some theory that suggests that with increasing CO2 you get some sort of saturation effect or asymptote in terms of the amount of heat increase you get with increasing CO2 above a certain level. But who's model is predicting that?
conrad
3 Mar 10 at 8:57 pm
Surely it makes more sense to start at a time point when the amount of CO2 humans started belching out into the atmosphere began to increase dramatically (as can be seen in the linked paper).
Why? You could end up getting the wrong signals. You’d want a long period of them before industrialization too so that you’re not confusing correlation with causation.
JC
3 Mar 10 at 9:14 pm
oops of time.
JC
3 Mar 10 at 9:14 pm
JC,
I’m sure those points are interesting also — however, isn’t it the case that we miss a lot of data from those years about other things that were going on? It’s not like people were keeping great records of solar flares and so on in 20543 BC.
conrad
3 Mar 10 at 9:53 pm
No argument there, Conrad. However it is problematic looking to draw observational conclusions from just that period you mention.
JC
3 Mar 10 at 9:59 pm
Samuel J, do the math. With iid innovations, the standard deviation of the monthly series will be sqrt(12) times that of the annual sums, and since sqqrt(12)=sqrt(180)/sqrt(15), the power of the two tests will be identical. With more complex lag structures, it could go either way. But you don’t get substantially more (or less) power by going from annual to monthly.
John Quiggin
3 Mar 10 at 10:51 pm
“no reason to start at 1988 either.”
The best thing to do is to use all the data, fit a model supported by theory, test hypotheses on the parameters, and then check your predictions against future outcomes. Hmm. I wonder if anyone might have done that already.
John Quiggin
3 Mar 10 at 10:53 pm
SHRILL INDEED
The question to Jones, Roskam’s misrepresentation of Jones’ answer on Q&A, the frenetic repetition of this material on right wing blogs, and Sinc’s posts on same look like another example of the creation and diffusion of AGW denialist talking points.
In this case, the talking point has clearly been exposed for what it is, albeit only after all manner of red-herring rationalisations created by AGW denialists to justify it have been debunked too. Presumably this talking point will now be dropped for a while, probably only to resurface again later – perhaps in a slightly different form – in an attempt to further befuddle the debate. Such is the pattern of AGW denialism.
As Quiggin commented (here):
Tom N.
4 Mar 10 at 9:21 am
Hmm. I wonder if anyone might have done that already.
.
Yes, and the predictions were not supported. To quote one prominent climate scientist, “The fact is that we can’t account for the lack of warming at the moment and it is a travesty that we can’t.”
daddy dave
4 Mar 10 at 9:22 am
I absolutely applaud any clarification and discourse on the analysis of climatology and the economics of climate change. If the results stand that there is significant and damaging AGW that requires mitigation, so be it.
I get similarly disappointed when Senator Wong has been caught out with some rather rubbery information on sea level rises and the drought in the MDB. Ditto for the hockey stick and ignoring the MWP.
Semi Regular Libertarian
4 Mar 10 at 9:50 am
“Yes, and the predictions were not supported.”
Perhaps you need to update your facts to include 2009 (e.g., http://www.giss.nasa.gov/research/news/20100121/). In addition, if 2010 and 2011 happen to be hot also, will you change your opinion?
conrad
4 Mar 10 at 10:00 am
Don’t want to be horrible here Tom, but you’re cherry picking. There are valid and interesting debates to had over the statistical analysis. John is contributing to that debate – you are not.
Sinclair Davidson
4 Mar 10 at 10:20 am
In addition, if 2010 and 2011 happen to be hot also, will you change your opinion?
.
No, because the fact that the earth is experiencing modest warming is a profoundly uninteresting finding.
.
The only reason people are paying attention is because of complex iterative models that predict all kinds of crazy stuff down the road. Having seen plenty of such modelling taking place in various labs at various organisations, I find that kind of evidence unconvincing, to say the least. If you factor in the evangelical zeal of many scientists involved in the debate, the doubt is compounded.
daddy dave
4 Mar 10 at 10:21 am
Warning per se does not provide support for the AGW hypothesis. We already know that the planet warms and cools over time anyway – although the process is not well undertood.
Sinclair Davidson
4 Mar 10 at 10:28 am
Already the phrase “pro-science Republican” is seen as an oxymoron, and the same is coming to be true of the right throughout the English-speaking world.
.
Nice meme, there.
The vilifcation of sceptics continues apace.
daddy dave
4 Mar 10 at 10:37 am
“No, because the fact that the earth is experiencing modest warming is a profoundly uninteresting finding”
.
Actually, whether the warming happens to be man made or not I imagine is profoundly interesting to many people (not least of those who happy to live in Bangladesh).
.
Also, if you’re not willing to change your opinion based on data, then it pretty much shows JQ is correct — you are happy to live in a world that lacks reality.
conrad
4 Mar 10 at 10:44 am
So Conrad,
What is the warming or cause of warming since 900?
Semi Regular Libertarian
4 Mar 10 at 10:51 am
Conrad – there are two hypotheses at work. First the planet is getting warmer. Second that is occuring due to human activity. Untangling those two hypotheses is difficult. You seem to be suggesting that demonstrating the first is enough.
Just for the record, let me just say that I am not a republican – I voted ‘no’.
Sinclair Davidson
4 Mar 10 at 11:02 am
Actually, whether the warming happens to be man made or not I imagine is profoundly interesting to many people
.
Yes, that’s interesting. However modest, small warming – which is what has occurred – in and of itself is not interesting.
.
Also, if you’re not willing to change your opinion based on data, then it pretty much shows JQ is correct — you are happy to live in a world that lacks reality.
.
There’s an old scientific axiom: Extraordinary claims require extraordinary evidence. Incremental shifts in temperature are not extraordinary evidence. I was a skeptic before this “lack of warming”; and had fully factored in further temperature increases into my scepticism. It was simply icing on the cake that the predictions failed so quickly.
daddy dave
4 Mar 10 at 11:24 am
“Yes, that’s interesting. However modest, small warming – which is what has occurred – in and of itself is not interesting.”
.
That’s not true. Why do you think they are building desalination plants all over Australia? Why do we have to pay farmers drought relief every 4 years instead of 6?
.
“There’s an old scientific axiom: Extraordinary claims require extraordinary evidence. Incremental shifts in temperature are not extraordinary evidence”
.
I have about the same evidence level of evidence for myself growing old. That doesn’t I won’t die of it.
.
“It was simply icing on the cake that the predictions failed so quickly.”
.
What predictions are you talking about? As per the previous NASA data, 2009 was hottest year in the Southern hemisphere ever recorded and the second hottest world wide. Are you just denying that data?
conrad
4 Mar 10 at 11:31 am
Daddy Dave, you might want to reflect on the point about the selection of 1995 as a starting point for this particular meme – someone is lying to you, and it isn’t the climate scientists.
As regards ‘extraordinary claims’, the science of the greenhouse effect is more than 100 years old and well established, so it would be extraordinary if the data didn’t reflect it.
John Quiggin
4 Mar 10 at 11:34 am
Why do you think they are building desalination plants all over Australia?
.
Because Australia has a water shortage. But this isn’t new, unless you think the Simpson Desert was a lush rainforest a couple of years ago.
.
Why do we have to pay farmers drought relief every 4 years instead of 6?
.
Again, that’s got nothing to do with global warming. Anyone who tells you otherwise is lying.
.
I have about the same evidence level of evidence for myself growing old.
.
No you don’t. You have the evidence of observation of countless individuals who grew old and died. I await your evidence from all the other planets that we’ve watched their ecosystems collapse from global warming.
.
Are you just denying that data?
I withhold judgement on the data until the heat island question is sorted out, as should everyone.
daddy dave
4 Mar 10 at 11:43 am
Sinc said:
As I replied to you previously, I do not mind the use of pejoratives, provided it is deserved. However, my comment clearly did not attempt to point out the specific flaws in your latest attempt to promulgate and then defend this talking point, as Q had already done that more than adequately. I simply put your efforts in a context.
Does that constitute ‘cherry picking’? Had I ignored your current post and instead extracted one of your posts from some time ago and brought it forward here to make my case, that would support your accusation. But what I did was respond to the current series of posts by you and say that they “look like another example” of a particular pattern. Morevoer, I can point to numerous other examples of this same pattern of AGW denialism, documented at sites such as Quiggin and Deltoid.
Sinc, you really are starting to look like the Black Knight on this topic. As I said, “shrill indeed”.
____________
DD said
Actually, from my reading, Quiggin was vilifying denialists; he has no problem with true sceptics. The problem is that most AGW ‘doubters’ appear to fall into the former group.
Tom N.
4 Mar 10 at 11:53 am
Daddy Dave, you might want to reflect on the point about the selection of 1995 as a starting point for this particular meme
.
John, I accept your point about that. My concern is that leaves the question of increasing” “accelerating” “flattening” or “decreasing” unanswerable, except in a hundred years from now. The skeptics are arguing that warming has leveled off; it seems that rather than disproving them, you’ve merely argued that the question can’t be answered.
daddy dave
4 Mar 10 at 12:04 pm
Tom – you’re trying very hard to shut down debate and when confronted you try to slither out with very fine distinctions. That is your perogative, as long as you don’t disturb the peace we tolerate all sorts here – but don’t get upset when we call ‘bullshit’. Now the argument isn’t about me, the argument is about how to interpret a comment made by Phil Jones. Given all the fuss, this is an important comment and getting the interpretation correct is important too. Now you can vilify the debaters or you can contribute to the debate.
Sinclair Davidson
4 Mar 10 at 12:17 pm
As I said, “shrill indeed”.
.
Tom, quoting yourself is dangerous, since it risks the perception of what is colourfully known, in the Australian vernacular, as being “up yourself.” It should be saved where you’ve mounted a particularly cogent argument that would be effortful and tedious to have to mount again.
.
“Shrill indeed” isn’t that.
daddy dave
4 Mar 10 at 12:45 pm
If I understand Quiggins point, it is that Jones has somehow been tricked into answering the question of the significance of a 15 year temperature trend, when the “correct” answer is “15 years is too short a time to calculate the significance”.
Is this a fair interpretation? If I recall, JQ talks about Jones being asked dishonest questions, whatever that means.
When asked, Jones said the trend was “quite close” to the (95%) significance level. It was “only just” not significant. How can that be interpreted any other way than “not significant at the 95% confidence interval”?
I’m not a statistician, but wouldn’t significance in a shorter trend simply require a steeper slope?
In regards to significance levels, I recall from my long ago stats course that the 95% significance is a weak level, and the 99% level is required for reasonable certainty that the null hypothesis is rejected. At least that’s what I remember as applying to clinical research on animals.
Yet Sinclair was talking about the 90% level earlier.
Is my memory faulty, or is it different in climatology?
Paul Williams
4 Mar 10 at 1:22 pm
In fact, Sinc, as an AGW policy sceptic, I would very much welcome serious debate and commentary that seeks the truth w.r.t the existence of AGW and, then, what if anything to do about it. I just don’t see much evidence of this approach among many AGW doubters.
In this context, the distinction between ‘sceptics’ and ‘denialists’ is in fact quite important. This is because the credibility of the position I hold is undermined by the behaviour of AGW denialists, who make it too easy for AGW policy activitists to paint all opposition as loony right wing reactionary stuff.
Tom N.
4 Mar 10 at 3:16 pm
In this context, the distinction between ’sceptics’ and ‘denialists’ is in fact quite important
.
That’s just your lame excuse to use abuse instead of reasoned debate. I’d like to hear an AGW proponent name someone who they think is a “skeptic” but not a “denialist”. As far as I can see it’s an empty category… if you doubt the theory, according to the AGW proponents, you’re a “denialist”.
daddy dave
4 Mar 10 at 3:46 pm
What is the difference?
Semi Regular Libertarian
4 Mar 10 at 3:55 pm
“I’m not a statistician, but wouldn’t significance in a shorter trend simply require a steeper slope?”
.
Yes, if the variance stays the same. But you wouldn’t want to ignore small trends in many instances, which is why you need more time points (which exist in any case). I age slowly and at an essentially undetectable amount from day to day, for example, but that doesn’t mean I’m not aging.
.
“In regards to significance levels, I recall from my long ago stats course that the 95% significance is a weak level, and the 99% level is required for reasonable certainty that the null hypothesis is rejected. At least that’s what I remember as applying to clinical research on animals.”
.
The level you take is essentially arbitrary, but in many fields people like 95% significance. However, if you don’t have much time to act, then you might accept a lower level, or if the possible outcome was excessively bad, then you would accept almost any level, which is why you get insurance, or body parts removed if you have cancer.
conrad
4 Mar 10 at 4:23 pm
According to Wiki:
scepticism is loosely used to denote any questioning attitude, or some degree of doubt regarding claims that are elsewhere taken for granted.scepticism: a personal disposition toward doubt or incredulity of facts, persons, or institutions; whereas
denialism is choosing to deny reality as a way to avoid an uncomfortable truth.
Tom N.
4 Mar 10 at 4:24 pm
Tom, you’re avoiding the question, “name someone who they think is a “skeptic” but not a “denialist”.”
Anyone at all. Is that category non-empty.
daddy dave
4 Mar 10 at 4:31 pm
who is ‘they’ DD?
Tom N is a policy skeptic, that not enough for you? is this some sort of inquisition on the Right now? John Humphreys is another person who is a policy skeptic but not a denialist i.e. he believes AGW is occuring.
jtfsoon
4 Mar 10 at 4:33 pm
I have a presentation that I gave last year where I set out a giant decision tree and looked at various costs and the like as we move through that tree. I can send you the slides if you like, I could also do a giant cut and paste here but it would look terrible. Let me know.
Sinclair Davidson
4 Mar 10 at 4:34 pm
Given how foundational the physics and thermodynamics behind AGW is I do doubt the fundamental intellectual sincerity of those who assert that it’s all a ‘conspiracy’ to get more public funding and make a mountain out of a molehill of certain anomalous results here and there and who proclaim the Birdian idea of the ‘end of peer review’ Such people are aptly classified as denialists.
I agree there are those intermediate who are genuine sceptics e.g. Pielke and Lindzen.
See this if you deny there is an attempt at a nuanced distinction among those who accept the mainstream science
http://www.foreignpolicy.com/articles/2010/02/25/the_fp_guide_to_climate_skeptics
jtfsoon
4 Mar 10 at 4:41 pm
Jason, “they” in the original context refers to anyone who uses the word “denialist”. I should have rephrased to make it clearer. In short, it seems to me like all skeptics are being labeled ‘denialists’ who are contrasted to the more reasoned “skeptics” who they should emulate, but who nobody can actually name.
.
a policy skeptic but not a denialist i.e. he believes AGW is occuring.
.
Okay. That answers my question.
Anyone who doubts catastrophic AGW is still a “denialist”. However being a mere policy skeptic (but accept catastrophic AGW) promotes you to the category of “skeptic.”
daddy dave
4 Mar 10 at 4:43 pm
Thanks for that extra link Jason, you posted it as I was writing. But I’d make the point that the word “denialist” is of no use at all. Someone like Tom N turns up, makes no points of his own but declares the arguments he sees to be ‘denialist’ arguments rather than ‘skeptical’ arguments. It’s simply name calling, nothing more. Not even backed up by reasons for the judgement.
That’s what you might call a “non-significant contribution” to the debate.
daddy dave
4 Mar 10 at 4:48 pm
TomN
How would you categorize Monckton?
JC
4 Mar 10 at 5:25 pm
As I pointed out earlier, rather than duplicate the work of Quiggin, my ‘value add’ to this thread was to put the talking point promulgated and then defended by Sinc into a broader context. As mentioned, there is no shortage of expose’s of other talking point exercises of this nature available inter alia at the Deltoid and Quiggin blogsites.
DD says that categorising the creation and diffusion of talking points as exercises in denialism is ‘simply name calling’. That may be true but, as I have previously observed (here), sometimes a pejorative label is well deserved, and serves a useful purpose in public debate. One suspects that DD’s sensitivity on this matter is that he senses that the cap fits.
Tom N.
4 Mar 10 at 6:31 pm
John,
“The glaring problem is that there is no reason to cut off the data at 1995 *except* that this as far back as you can go without getting a statistically significant trend. The trend since, say, 1988, is clearly significant.”
Obviously you can pick any start and end points you like, and get different trends.
Monckton points out the “accelerating trend” trick used by the IPCC.
It’s also obvious by eyeballing the chart that the trend has flattened off, hence Jones’ admission that the trend since 1995 is “only just” not significant at the 95% confidence level.
conrad, thanks.
Are you saying the confidence level as calculated for studies written for peer-reviewed publication, is dependent on what is being studied. And that climatology is happy with a 1 in 10 (or 1 in 20) chance that the null hypothesis is not disproved?
Paul Williams
4 Mar 10 at 6:50 pm
I mean,
the confidence level that is acceptable for publication
Paul Williams
4 Mar 10 at 6:52 pm
Jason I haven’t read all the FP article you linked to, but when they discuss criticism of McIntyre and McKitrick without mentioning the Wegman report in their defence then you have to wonder about the quality of the piece.
Pedro
4 Mar 10 at 7:12 pm
daddy dave — I don’t think we are facing *catastrophic* global warming. I hold my views on the science lightly because I think we’re still on a steep learning curve in this relatively young science… but based on my understanding of current information I believe that since around 1970 we have seen an upward trend in temperatures, in part caused by human emissions of greenhouse gases. I find the main estimates of IPCC estimates quite believable (2-4 degrees).
However, I do think many in the media have exaggerated the real science into a scare campaign. And I haven’t seen any AGW policies that would pass a benefit-cost analysis, except perhaps a well structured tax-swap. And on the science, I think there is a chance that the sensitivity has been over-estimated and that we may be facing the bottom of the IPCC estimates (though I think it reasonable to do policy analysis on the assumption of 2-4 degrees).
As I wrote long ago, I split the protagonists in this debate into four categories… denialists (just say “no”), skeptics (follow the science, but not scared), mainstream (follow the science and a bit scared) and alarmists (the sky is falling). Unfortunately, the denialists and alarmists get most of the airplay because they are the most dramatic. They both also have the unfortunate habit of ignoring the sensible middle.
I suggest some other skeptics include Bjorn Lomborg and Patrick Michaels. Other skeptics in the blogosphere would include Jason Soon, Terje Petersen and JC. And perhaps TomN, though I don’t really know his position. Some non-alarmed mainstreamers would include JohnZ and Jarrah.
I admit that some people on “my” side of politics are denialists, but I would also accuse plenty of people on the “other” side of politics of being alarmists. I can’t see much point in engaging either group because their conclusions are fixed.
John Humphreys
4 Mar 10 at 8:16 pm
“I hold my views on the science lightly because I think we’re still on a steep learning curve in this relatively young science…”
I used to hold that view. Don’t you feel cheated when you try to be honest, accept the science and give an educated economic response, given that is your forte, only to find the climate change Minister is making some blatant whoppers (MDB drought, 2 metre sea level rise predicted by IPCC, etc.)?
Semi Regular Libertarian
4 Mar 10 at 8:25 pm
I can’t see much point in engaging either group because their conclusions are fixed.
.
There’s always point in engaging with people, no matter how fixed their views seem to be.
Quite apart from anything else there are always bystanders to these debates, who will silently judge the merits.
daddy dave
4 Mar 10 at 8:35 pm
“Obviously you can pick any start and end points you like, and get different trends.”
So, the sensible thing to do is to fit your model to the entire data set, then make projections, as Hansen et al have done since the 1980s. Contrary to some claims above, the estimates made in the 1980s fir the subsequent data very well.
“Monckton points out the “accelerating trend” trick used by the IPCC.”
Since Monckton is a loon who believes the whole thing is a plot to institute a communist world government (his words, not mine), I would not give any weight to any claim he makes. If you can actually do the statistical analysis yourself, that’s fine, but then there’s no reason to rely on Monckton.
“It’s also obvious by eyeballing the chart that the trend has flattened off, hence Jones’ admission that the trend since 1995 is “only just” not significant at the 95% confidence level.”
The whole point of hypothesis testing is that patterns that look convincing to the human eye often turn out to be nothing more than random chance. As Jones says later, attempts to fit a cooling trend to short bits of the data completely fail tests of significance. There’s no evidence to reject (in either direction) the null hypothesis of a constant upward trend
John Quiggin
4 Mar 10 at 10:07 pm
TomN
I watched the Monckton Lambert debate and I cannot really place Monckton in the denialist camp as some alarmists have done in the past.
I was quite surprised as I didn’t really know much about Monckotn other than what alarmists said about him.
He believes in the lab science and the only real disagreement between the two was the rate of change and the level of temp rise CO2 concentrations will cause.
That to me doesn’t sound like a denialist. Does it?
JC
4 Mar 10 at 11:03 pm
That’s not quite right. Jones rejects that null in his answer to question B, saying only just. When I guesstimate his analysis I get a p-value just over 0.05, so I’m inclined to believe Jones on that score. This raises the question as to why anyone would think that just over 0.05 is ‘bad’ and just under 0.05 is ‘good’. It also raises the question why he didn’t just say, yes at 95% but no at 90% – people might grumble a bit but there’d really be nothing to say. Remember Jones volunteers the significance level the question doesn’t.
Sinclair Davidson
4 Mar 10 at 11:06 pm
John Humphreys:
I’m not that sceptical actually. Think we have a pretty big problem longer term that can be fixed by allowing and encouraging nuclear power to take its proper place.
Funnily enough the person who has changed my opinion the most about this issue is someone with perfect green credentials and that is Barry Brook.
He’s a young and very bright scientist who applies straight up and down logic to this issue.
He thinks we have a big problem but the way to solve it is not by making electricity ruinously expensive simply wrecking our economy and more ominously, the economies of the poor countries.
He thinks the least damaging way to de-carbonize is to head to advanced nuclear energy that with time should see the price of energy drop below that of carbon which we will begin to see through economic scaling.
Reason for optimism on that score was that the recent contract awarded to the Korean engineering firm was 30% below the American bid using the same Westinghouse reactor technology. Plant installation over the next 30 years should collapse if this technology isn’t hobbled.
JC
4 Mar 10 at 11:19 pm
I’d say I’m closer to mainstream than sceptic too. I don’t think a rise of 2-4 degrees with its possible repercussions not to mention ocean acidification is anything to scoff at. At the very least I detest warm weather and I like my seafood.
Jason Soon
4 Mar 10 at 11:23 pm
Currently the physical evidence (ice loss as measured by GRACE satellites) points to the upper end of projections so arguing about nuclear or null points may be of no real significance.
rog
4 Mar 10 at 11:50 pm
JC – I have loads of time for Barry Brook.
TerjeP (say tay-a)
4 Mar 10 at 11:56 pm
I pretty much agree with this breakdown. I think that the sceptics, as defined above by John, don’t mind the science debate festering on because they’re not scared, whilst those that are scared find the festering science debate infuriating if it delays a rush to the medicine cabinet for an instant cure.
TerjeP (say tay-a)
5 Mar 10 at 12:00 am
Why would arguing about nuclear be of no real significance, Rog?
What’s your wonderful solution? Do you have a better idea. Heard of one?
The US, China, India, Brazil, Mexico, The EU, Vietnam, Indonesia, Russia to name but a few places, are not introducing an ETS of it’s derivations.
In other words despite all the hoopla out of the Copenhagen haj there is no global commitment to arrive at any global mutually agreed decision to raise the price of carbon based energy.
In fact reading the tea leaves, the obama administration is going to push towards nuclear power.
Please explain how exactly we can achieve global reductions under the present arrangements?
Be also aware that even if Rudd were successful in hobbling our economy with that abomination of a policy how would that exactly lower global emissions in any material way when the rest of the world is not following Rudd?
Lastly why are you supporting a policy that would lower energy consumption when you want to actually increase it around the world.
JC
5 Mar 10 at 12:07 am
JC – I have loads of time for Barry Brook.
Absolutely. He’s even convinced me about environmentalism.
He’s convinced me that it shouldn’t be ideological in the slightest.
This is truly one of the best threads I’ve ever read, paying particular attention to Peter Lang who makes some of the best comments around.
http://bravenewclimate.com/2010/01/09/emission-cuts-realities/
JC
5 Mar 10 at 12:13 am
Because
1. nuclear is a realtively old technology that still requires massive govt subsidy to exist
2. despite the best wills in the world it still takes years to complete a plant
3. waste disposal is a massive problem still not yet resolved
rog
5 Mar 10 at 12:16 am
1. Wind is older and needs bigger subsidies.
2. And they last a long time.
3. It is a small problem. Coal fired power plants put out unmanaged nuclear waste and we seem to live with it. Managed nuclear waste seems like a big improvement.
TerjeP (say tay-a)
5 Mar 10 at 12:21 am
1. Coal is proven cheap technology.
2. We have stacks of it.
3. No problem.
Infidel Tiger
5 Mar 10 at 12:30 am
1. nuclear is a realtively old technology that still requires massive govt subsidy to exist
Are you nuts? It’s 50 years old and in the time span of energy production technology humans have witnessed it’s a baby. Taking into account the stop/go periods the technology has experienced it’s still in the womb.
And since when has old or new meant anything in terms of economic efficiency. We still use wheels and that technology is 1,000 of years old.
Add also what Terje said about the money being thrown at the two subsidy whores, which I believe won’t ever work, as it results in the indivisibility of costs through economies of scale.
Ask yourself why we don’t have hundreds of coal fired plants whereas we would need the equivalent of thousand of windmills and solar plants. Indivisibility has serious economic consequences.
2. despite the best wills in the world it still takes years to complete a plant
See what Teje said.
I would also add that is a bullshit argument you make as France took 15 years to go nuke, straight after the 70′s oil shock.
3. waste disposal is a massive problem still not yet resolved
Please. Waste disposal is not a massive problem especially with the introduction of large scale 3rd generation breeders reactors that consume most of the fuel and leave little waste.
JC
5 Mar 10 at 12:43 am
tiger:
Given a chance I have no doubt nuke could produce energy at a cheaper cost than coal. It could be dirt cheap.
The cost of nuke is 90% plant and 10% fuel. I see the cost of plants collapsing as more get built and economies of scale set in, as we begin to make the process much more methodical.
If humans are good at anything, as we progress though the industrial technological age, it’s to make relatively expensive things cheaper from the tech point.. see cars for example.
The more “mass production” and standarization of the nuke plants we see the cheaper they would become.
In fact the capital intensity and the ability to centralize large scale energy production from large units suggests that nuke could beat coal in costs.
You don’t require many people to operate a plant.
JC
5 Mar 10 at 12:52 am
Sinclair: Using monthly data you end up with 15 x 12 = 180 data points.
Because you don’t really have 180 data points with monthly data, for two reasons:-
1. There’s orbital variation, the Earth has an elliptical orbit which means it’s close to the Sun in January than July
2. There’s hemispheric variation. There is a greater mass of water in the Southern Oceans than the Northern, and conversely a greater mass of vegetation in the Northern Hemisphere than the Southern.
So you have to use annual figures which eliminate the annual cycle.
JM
5 Mar 10 at 2:50 am
So what collect monthly figures at all, JM.
JC
5 Mar 10 at 2:59 am
JM – that’s bullshit. 15*12 = 180 every day of the week, every week of the year. The question you want to ask is whether 180 monthly observations is a sufficient sample. I have already commented on that.
Sinclair Davidson
5 Mar 10 at 7:44 am
Believe what you will
rog
5 Mar 10 at 7:49 am
Read this
rog
5 Mar 10 at 7:56 am
and has been declining for several years.
.
Isn’t that because of the anti-nuke movement across the West?
If so, it’s very clever. First, you lobby to stop development of nukes. Then you point to the fact that no nukes have been built lately as evidence that it’s not feasible.
daddy dave
5 Mar 10 at 7:58 am
who is ‘they’ DD? Tom N is a policy skeptic, that not enough for you? is this some sort of inquisition on the Right now? John Humphreys is another person who is a policy skeptic but not a denialist i.e. he believes AGW is occuring.
DD asked a fair question. If Tom N proposes two categories and he’s unable to provide examples for one of them we’re entitled to think his use of this purported distinction is disingenuous. By the way, being a policy sceptic doesn’t count, I think, in this instances since what were concerned with here is people’s attitude towards the science, not the economics.
dover_beach
5 Mar 10 at 9:28 am
Sinclair, the claim about 180 observations is absolutely wrong, as I’ve already explained. Let me try again.
It’s well known that, with a sample size of 1000, you need a 3 per cent shift in opinion polling numbers on a question like “Do you approve of X’s performance” to reject the null hypothesis. Let’s suppose that a given poll showed a 5 per cent shift, so we would assume a genuine change and not just sample variation.
Suppose I presented the same data to you, but as 50 data points, each representing the average for 20 of the respondents (say from the subsample interviewed in particular suburbs). Trivially, the change will still be 5 per cent – it’s the same data. Would you now say that the change was statistically insignificant since there are fewer observations?
John Quiggin
5 Mar 10 at 9:37 am
John – I think we’re arguing at each other and not with each other.
I ran a regression with a trend variable post 1995 using monthly data and looked at the coefficient and p-value. The coefficient was almost the same as that Jones reports and the p-value (0.0512) is close to what Jones suggests.* I do the same thing with annual data. As I expect the standard error is larger with the annual data and the p-value higher (0.2446 with an AR(1) – what I did before and 0.0896 without the AR(1)). The coefficient is smaller but still pretty close to what Jones suggests. I would expect that increasing the frequency of the observations would give a better standard error and a smaller standard error (we know as the sample size increases that the probability of rejecting the null increases anyway, but we’re nowhere near that problem with 15 years of monthly data). Now maybe he has used annual data – that’s pretty dumb given that he has monthly data to play with.
*Using the mechanical rule of thumb that Jones and the BBC interviewer is relying on, we are arguing over a difference of 0.0012 in the p-value.
Sinclair Davidson
5 Mar 10 at 10:21 am
Sinclair
“It also raises the question why he didn’t just say, yes at 95% but not at 90%”
This puzzles me. My faint recollection of statistics is that 95% is a weak level of confidence, and 99% was really required, especially for publication.
Is it different in climatology?
John
“So, the sensible thing to do is to fit your model to the entire data set”
So you agree with Bob Carter. Just kidding.
The start and end points matter, and there are trends of varying length, so while you can continue to insist that 1988 is the best start point, since that fits your argument, there is an indisputable recent flattening. (And that assumes the temperature data is uncorrupted).
“Monckton is a loon ….”
No, he’s not.
Do you have a comment about the point I raised?
“The whole point of hypothesis testing is that patterns that look convincing to the human eye often turn out to be nothing more than random chance.”
Yes, I know. However, Jones himself, in his BBC Q&A, assigns significance to periods of 21 and 24 years, so perhaps he does not agree with you that 30 years is the minimum necessary. I pointed this out to you on your blog, but you did not comment on it.
Paul Williams
5 Mar 10 at 10:25 am
Ok, that’s embarrassing. How do I put bold and italics in?
[Instead of using square brackets use the greater than and less than signs. Sinc]
Paul Williams
5 Mar 10 at 10:26 am
Paul, because he is not used to being interviewed?
Butterfield, Bloomfield & Bishop
5 Mar 10 at 10:45 am
Paul – the significance levels are somewhat arbitrary and depend on the trade-off between Type I and Type II errors. Generally speaking 99% is better than 95% is better than 90%. Most people report the p-values or standard errors and make an argument why they think the coefficient is important. Less than 90% is not going to be accepted. But as I said above the p-value I get is 0.0512 (so the significance is 100% – 5.12% = 94.88%).
Ziliak and McCloskey talk about the rule of three v the rule of two and Stephen Stigler (son of the great George) has described the history of this too.
Sinclair Davidson
5 Mar 10 at 10:56 am
Paul it’s standard html. That is,
<i>i to start italics; then </i> to end them.
Ther’es also <blockquote>i to start a blockquote; then </blockquote> to end it, and
<strong>i to start bold then </strong> to end it.
daddy dave
5 Mar 10 at 11:02 am
oops… why didn’t that work?
daddy dave
5 Mar 10 at 11:02 am
Insignificant warming trends: why 1995 was chosen
http://motls.blogspot.com/2010/03/insignificant-warming-trends-why-1995.html
dover_beach
5 Mar 10 at 12:21 pm
I use this site for HTML testing, I hope it helps:
http://www.play-hookey.com/htmltest/
Capitalist Piggy
5 Mar 10 at 1:12 pm
Sinclair
Thanks, that gives me something to work on re significance levels.
B,B&B,
That’s ridiculous. The BBC Q&A was hardly an ambush. There was follow up from Jones/UEA. The first question includes the table with four periods (lengths, 21, 31, 24, 35 years respectively), trends, and “yes” in the statistically significant column.
http://news.bbc.co.uk/2/hi/science/nature/8511670.stm
Thanks for the help with bolding.
Paul Williams
5 Mar 10 at 1:32 pm
Paul,
Did I say it was ambush?
If you have little experience with the media you can forget the obvious and say things you didn’t really want to say or thought relevant
Butterfield, Bloomfield & Bishop
5 Mar 10 at 1:35 pm
I’m wrapping up now, but I’ll just offer an invitation to a zero stakes bet. Who wants to bet that, if we add 2010 into the data set used by Jones, that the null hypothesis will not be rejected?
John Quiggin
5 Mar 10 at 1:47 pm
BBB,
He brought along a table with those figures on them, FFS.
Actually, I believe it was an email exchange, with follow up. (He he)
He said exactly what he wanted to say.
Paul Williams
5 Mar 10 at 1:55 pm
Actually, I believe it was an email exchange, with follow up…He said exactly what he wanted to say.
Yes, Paul is right.
dover_beach
5 Mar 10 at 2:02 pm
Paul, are you that naive?
Why do you think execs get training with the media?
you can have tables up to your armpits and still get confused after all people here are confused about the statistics until John helped them out
Butterfield, Bloomfield & Bishop
5 Mar 10 at 2:03 pm
Paul, are you that naive?
Jones has been interviewed many other times; he is hardly a media virgin.
you can have tables up to your armpits and still get confused
If he can get ‘confused’ while answering questions provided to him via email then he shouldn’t be heading the CRU.
after all people here are confused about the statistics until John helped them out
John, I believe, is wrong, although I’m happy to change my mind if convinced otherwise. Motl says, in the comments to an above linked post, in relation to the monthly/ yearly problem among other things:
dover_beach
5 Mar 10 at 2:29 pm
I’ll re-run the regression with the addition 12 months and post the result when the data become available.
Sinclair Davidson
5 Mar 10 at 3:49 pm
BBB, perhaps Jones should consult with JQ then.
Paul Williams
5 Mar 10 at 5:19 pm
BBB/Homer, stop ruining a good thread with crap.
JC
5 Mar 10 at 5:44 pm