It’s too bad, I really used to enjoy reading Phil Plait.
I still do, as long as he sticks to astronomy. I just try to ignore the rest.
I used to be a regular at the BAUT forum until I started questioning AGW. I finally gave up the site which was a shame. It was a good place to discuss space but questioning AGW went over about as well as questioning the Koran in Saudi Arabia.
I’ve become wary of science bloggers who excessively use the exclamation point. No mere laughing-at-your-own-joke here. It’s like they’re trying to engage jaded middle-schoolers by channeling the ghost of Carl Sagan — the exclamation points are ritual candles.
10 years is too short a period to assert with reasonable confidence what the global temperature trend is, so first off, the 16 scientist are wrong to make such a claim, Plait, in his eagerness to argue that temperatures are still rising went on to wrongly argue that extrapolation of the historic trend was enough to argue that this proved the trend was continuing.
Briggs was wrong to claim that because of Plait’s error, the 16 must be right, which is BS logic.
Just to clarify, when I said “the 16 scientist are wrong to make such a claim” I’m referring to their statement: “Perhaps the most inconvenient fact is the lack of global warming for well over 10 years now.”
Trenberth (et.al.) responded in today’s wsj. It’s entertaining.
And computer models have recently shown that during periods when there is a smaller increase of surface temperatures, warming is occurring elsewhere in the climate system, typically in the deep ocean.
Our models (the ones that don’t show temperatures flat over the last 15 years) show deep ocean warming. Beautiful.
In addition, there is very clear evidence that investing in the transition to a low-carbon economy will not only allow the world to avoid the worst risks of climate change, but could also drive decades of economic growth.
I can see him running it by an associate before sending it off. “You should add something about green jobs. Cover all the bases”. “Good idea”.
Their models may show open warming, but all the thermometers we’ve got monitoring the oceans sure don’t.
I was just getting ready to ask if putting a thermometer in the ocean was a technology that escapes our current primative state.
I wish he’d stay the hell out of politics–and political climatology–and stick with astronomy.
A few months ago on the Compuserve SCIMATH forum (I know, Compuserve = kerosene lanterns, etc. etc.) we had a discussion about the curious statistics underlying temperature trendlines. The situation is a lot more subtle than most people realize. If you scroll forward to the 18th post, and look at the attachment, you will see a depiction of the high and low temperature trends in Memphis. It’s curious that the slope in multiple consecutive 8-year intervals can exceed the average slope over the total period, which is the converse of the point that Plait made (that the slope over multiple consecutive short intervals can be less that the overall trend). It does raise the question of what exactly the meaning of a regression line is and why we ever believe it is predictive….
I don’t get your point, if you cherry-pick years temperatures above the trend line at the start of a short (~decade long) period and finish with years that have temperatures below the trend line, as denialists often do, you can easily show short term temperature trends well below the overall trend.
Conversely, if you cherry-pick years temperatures below the trend line at the start of a short (~decade long) period, and finish with years that have temperatures above the trend line, as no one does, you can easily show short term temperature increases well above the overall trend line.
if you cherry-pick years temperatures above the trend line at the start of a short (~decade long) period and finish with years that have temperatures below the trend line, as denialists often do
Emphasis mine. “Denialists”? Really?
Is that supposed to be a synonym for skeptics (i.e., people actually interested in doing science)? Is the next step, after having declaring us heretics, to load us into the box cars? Or is it just to drum us out of decent society (the boxcars come later)?
I’m absolutely prepared to call genuine skeptics “skeptics”, but people who “cherry-pick short periods years with temperatures above the trend line at the start, and finish with years that have temperatures below the trend line” to argue that such periods demonstrate a reversal of the long term trend are lying to advance a political or ideological position.
Similarly if people were “cherry-pick[ing] years with temperatures below the trend line at the start of a short (~decade long) periods, and finishing with years that have temperatures above the trend line” would also, in my view be “lying to advance a political or ideological position” I’m happy to call such people “alarmists”.
Sometimes you have to call a spade a spade.
Or do you always think in terms of goodies and baddies? those who’re “skeptical” of AGW are always honest up standing citizens with the sun shining out of their a**, while those who’re accepting of AGW are always baddies – corrupt, dishonest, aiming to take over the world?
If that’s how you see things I think you should try to up date your perspective on the world, to something a little more in tune with reality than a really bad Bond movie.
Andrew, as a heads up, “denialist” was coined by CAGW believers to try to equate people who questioned their miserably incompetant use of data and statistics with Holocaust deniers, so they could tar good scientists and mathematicians as genocidal racists and win the argument based not on math and science, but on public perception conveyed in sound bites.
In engineering, if someone claimed to have a novel device such as a 200 mpg carberator that is constantly refuted by real world experiments, and they couldn’t actually produce a working model after 30 years, then resorted to a conspiracy theory involving big oil money, would you suck down their every press release as gospel? What about when then not so subtly tried to liken their detractors as Nazis? I would conclude that they’re a bunch of failing, desperate nutcases.
Andrew, “the topic of the post” is the application of statistics to temperature measurements. You’re stealing a base when you use “denialist” and “skeptic” (and “alarmists”) while casually including the acronym AGW. It’s a technique favored by Bob-1, and I think is arguably beneath you.
George, actually people who would cling to ideological reasons to deny evidence of AGW used to be called AGW deniers“, “denialist” was coined in a response the “link to Holocaust deniers” charge. To me the term is analogous to people who are in denial of having a drug or gambling addiction, no matter what evidence you put before them, they’re incapable of accepting the reality of the situation. They rationalize that it’s all a big conspiracy, that everyones against them etc, because their addiction to the present situation won’t let them look at things with any objectivity.
Well, thanks for jumping off the deep end there Andrew. How’s the water?
We’re addicted to the present situation, so we can’t look at things objectively? You’re mistaking the main take-away from the UEA correspondence. It’s not the indication of conspiracy, it’s the paranoia. The unmistakable attitude of “We have to protect our phoney baloney jobs here, gentlemen!” They’re the ones who are rationalizing that it’s all a big conspiracy. Big oil. Or something.
As far as clinging to ideology, the subject here is statistical methods and temperatures. “GW”. You’re the one reflexively adding the “A”. Why?
“It’s a technique favored by Bob-1”
What is the technique in question?
I don’t get your point, if you cherry-pick years temperatures above the trend line at the start of a short (~decade long) period and finish with years that have temperatures below the trend line, as denialists often do, you can easily show short term temperature trends well below the overall trend.
Conversely, if you cherry-pick years temperatures below the trend line at the start of a short (~decade long) period, and finish with years that have temperatures above the trend line, as no one does, you can easily show short term temperature increases well above the overall trend line.
I urge you to read the whole thread — it’s not very long. I was critiquing the observations of a pro-AGW blogger named David Kroodsma. Kroodsma performed an analysis through which he claimed to show that global warming was causing nighttime temperatures to rise faster than daytime temperatures. In my critique, I pointed out that (a) he was using a pretty short baseline to come to his conclusion, and (b) trends in min and max temperatures records are highly variable. Along the way I ran across a funny statistical paradox. I pointed out
Please note how two consecutive blocks can each have a higher regressed slope that the combined period, e.g. the low temp for 1992-1999 (7.96 F/dY) and 2000-2007 (4.92 F/dY) versus the slope for the low for 1992-2007 (1.48 F/dY).
This is sort of a time series version of Simpson’s paradox. In fact if you study the plot I made, you might notice a curious artifact: for every single pair of consecutive 8-year blocks, for both high and low temperatures, the end of the earlier block’s regression is higher than the beginning of the later block’s regression. I’ll let you ponder the significance of that observation, especially in light of the arguments that the most recent decade still regresses to a positive slope.
And somehow I think you missed the point that these time intervals were not “cherry-picked” to produce this effect. I had 60 years of daily hi/lo data for Memphis handy. I chose to segment it in 8 year blocks so that each block would start on the same day of the year and have exactly the same number of days. And consequently I chose to use the last 56 years — and all of this was done a priori. There was no cherry-picking. For the seven 8-year blocks, there were 4 in which the low temperature rose faster than the high temperature, and 3 in which the high temperature rose faster than the low temperature. So I am skeptical of the significance of Kroodsma’s result.
But nowhere in this critique do I accuse Kroodsma of bad faith or cherry-picking. (That is your invention.) I am merely observing a lack of competence, not goodwill. Like so many climatology researchers, Kroodsma does not bother to establish a null hypothesis or perform a significance test for his results. He simply makes a computation and publishes it with layers of interpretation that are not justified by his data.
if you cherry-pick years temperatures above the trend line at the start of a short (~decade long) period and finish with years that have temperatures below the trend line, as denialists often do,
Speaking of competence, I should point out that your understanding of this statistical phenomenon is less than adequate. It is NOT the case that in the Memphis data the blocks were selected to begin with temperatures on one side of the trend line and end with temperatures on the other side of the trend line. That is NOT what is going on here. If that were a sufficient condition to produce the effect you describe, there would be no way for every pair of blocks to show the effect I described:
for every single pair of consecutive 8-year blocks, for both high and low temperatures, the end of the earlier block’s regression is higher than the beginning of the later block’s regression.
Think about it.
OK, when I said I don’t get your point I was trying to square your comment with the topic of the post. I could see that you hadn’t cherry picked the 8 year blocks to engineer short term trends in the Memphis data.
When I said “you can easily show …” I should have said “people can easily show …” I wasn’t clear and my bad.
And you’re right, I don’t claim any expertise in statistics (mind you, people like Briggs can boast about their expertise as statisticians, but if they’re to lazy to actually do the math to support their argument – as Briggs didn’t) their opinion isn’t worth a damn.
Cheers.
(mind you, people like Briggs can boast about their expertise as statisticians, but if they’re to lazy to actually do the math to support their argument – as Briggs didn’t) their opinion isn’t worth a damn.
You know what? You’re right. And I should have commented on this earlier but got sidetracked defending my Memphis data. Although I don’t think Plait did himself any good with his post, I am not too impressed with Briggs’ arguments, either.
One of the treacherous aspects of regression analysis is the subtlety of the different kinds of uncertainty that are associated with the analysis. Strictly speaking, a least-squares regression should take into account the uncertainties attached to each and every data point. Briggs calls this “parameter uncertainty” or “error bars from the estimates” (which is ambiguous at best). I would call it the “input uncertainty” for the regression, or “pointwise uncertainty”.
Now, two points: (a) if the uncertainty is approximately the same for all the data points (say, 0.1 F across the board), then the results of the regression are not changed by including the uncertainty — the pointwise variance can then be pulled out of the chi-squared sum and becomes an overall factor. It is important to account for the “parameter uncertainty” mainly when there are outlying data points with large uncertainties, combined with a large number of better-behaved, better-known values. The other point is (b) for “well-behaved” datasets, i.e. those with approximately uniform uncertainties and a reasonably sound trendline (say R^2 ~ 0.8 or above), it is standard practice to perform the regression without accounting for pointwise uncertainty and let the R^2 value speak to the quality of the regression.
Now, what Briggs is getting at in his post — what he calls “prediction uncertainty” — is the fact that one of the results of regression analysis is a confidence band around the regression line. That is, the regression line as drawn is simply the most likely straight line that explains the data (“most likely” in the sense that it minimizes chi-squared). But there are other regression lines that are almost as good — some above, some below, some with higher slopes, some lower — and these form a probability distribution for the actual trend. The confidence bands around the trendline typical flare out at the lower and upper bounds of the x (date) range. If the trendline is uncertain enough, it is possible that the uncertainties at the highest and lowest dates overlap. That is the possibility Briggs is referring to when he writes, “I don’t know what the prediction uncertainty is for Plait’s picture. Neither does he. I’d be willing to bet it’s large enough so that we can’t tell with certainty greater than 90% whether temperatures in the 1940s were cooler than in the 2000s.”
Now, frankly, if you used other dates, like 1950s vs 2000s, I would bet against Briggs on that one, before running any analysis. But he carefully said “the 1940s” because he knows that global temperatures hit a local maximum in that decade before sliding down to a minimum in the 1970s. So he’s being a little disingenuous here. But likewise, so is Plait, who starts his graph — you guessed it — in the 1970s. If you did a regression from 1940 to 2011, I still think it likely that the 90% confidence band would exclude zero rise. But like any good bettor, I am going to run the numbers before I place my bet.
Everytime I see the term ‘denialist’, I have a sudden urge to order another case of ammunition. It’s more fun than buying indulgences from some modern medeval Pope wannabe.
It’s better than just waiting for the knock at the door.
It’s too bad, I really used to enjoy reading Phil Plait.
I still do, as long as he sticks to astronomy. I just try to ignore the rest.
I used to be a regular at the BAUT forum until I started questioning AGW. I finally gave up the site which was a shame. It was a good place to discuss space but questioning AGW went over about as well as questioning the Koran in Saudi Arabia.
I’ve become wary of science bloggers who excessively use the exclamation point. No mere laughing-at-your-own-joke here. It’s like they’re trying to engage jaded middle-schoolers by channeling the ghost of Carl Sagan — the exclamation points are ritual candles.
10 years is too short a period to assert with reasonable confidence what the global temperature trend is, so first off, the 16 scientist are wrong to make such a claim, Plait, in his eagerness to argue that temperatures are still rising went on to wrongly argue that extrapolation of the historic trend was enough to argue that this proved the trend was continuing.
Briggs was wrong to claim that because of Plait’s error, the 16 must be right, which is BS logic.
Just to clarify, when I said “the 16 scientist are wrong to make such a claim” I’m referring to their statement: “Perhaps the most inconvenient fact is the lack of global warming for well over 10 years now.”
Trenberth (et.al.) responded in today’s wsj. It’s entertaining.
And computer models have recently shown that during periods when there is a smaller increase of surface temperatures, warming is occurring elsewhere in the climate system, typically in the deep ocean.
Our models (the ones that don’t show temperatures flat over the last 15 years) show deep ocean warming. Beautiful.
In addition, there is very clear evidence that investing in the transition to a low-carbon economy will not only allow the world to avoid the worst risks of climate change, but could also drive decades of economic growth.
I can see him running it by an associate before sending it off. “You should add something about green jobs. Cover all the bases”. “Good idea”.
Their models may show open warming, but all the thermometers we’ve got monitoring the oceans sure don’t.
I was just getting ready to ask if putting a thermometer in the ocean was a technology that escapes our current primative state.
I wish he’d stay the hell out of politics–and political climatology–and stick with astronomy.
A few months ago on the Compuserve SCIMATH forum (I know, Compuserve = kerosene lanterns, etc. etc.) we had a discussion about the curious statistics underlying temperature trendlines. The situation is a lot more subtle than most people realize. If you scroll forward to the 18th post, and look at the attachment, you will see a depiction of the high and low temperature trends in Memphis. It’s curious that the slope in multiple consecutive 8-year intervals can exceed the average slope over the total period, which is the converse of the point that Plait made (that the slope over multiple consecutive short intervals can be less that the overall trend). It does raise the question of what exactly the meaning of a regression line is and why we ever believe it is predictive….
Sorry, I meant to link the 19th post….
I don’t get your point, if you cherry-pick years temperatures above the trend line at the start of a short (~decade long) period and finish with years that have temperatures below the trend line, as denialists often do, you can easily show short term temperature trends well below the overall trend.
Conversely, if you cherry-pick years temperatures below the trend line at the start of a short (~decade long) period, and finish with years that have temperatures above the trend line, as no one does, you can easily show short term temperature increases well above the overall trend line.
if you cherry-pick years temperatures above the trend line at the start of a short (~decade long) period and finish with years that have temperatures below the trend line, as denialists often do
Emphasis mine. “Denialists”? Really?
Is that supposed to be a synonym for skeptics (i.e., people actually interested in doing science)? Is the next step, after having declaring us heretics, to load us into the box cars? Or is it just to drum us out of decent society (the boxcars come later)?
I’m absolutely prepared to call genuine skeptics “skeptics”, but people who “cherry-pick short periods years with temperatures above the trend line at the start, and finish with years that have temperatures below the trend line” to argue that such periods demonstrate a reversal of the long term trend are lying to advance a political or ideological position.
Similarly if people were “cherry-pick[ing] years with temperatures below the trend line at the start of a short (~decade long) periods, and finishing with years that have temperatures above the trend line” would also, in my view be “lying to advance a political or ideological position” I’m happy to call such people “alarmists”.
Sometimes you have to call a spade a spade.
Or do you always think in terms of goodies and baddies? those who’re “skeptical” of AGW are always honest up standing citizens with the sun shining out of their a**, while those who’re accepting of AGW are always baddies – corrupt, dishonest, aiming to take over the world?
If that’s how you see things I think you should try to up date your perspective on the world, to something a little more in tune with reality than a really bad Bond movie.
Andrew, as a heads up, “denialist” was coined by CAGW believers to try to equate people who questioned their miserably incompetant use of data and statistics with Holocaust deniers, so they could tar good scientists and mathematicians as genocidal racists and win the argument based not on math and science, but on public perception conveyed in sound bites.
In engineering, if someone claimed to have a novel device such as a 200 mpg carberator that is constantly refuted by real world experiments, and they couldn’t actually produce a working model after 30 years, then resorted to a conspiracy theory involving big oil money, would you suck down their every press release as gospel? What about when then not so subtly tried to liken their detractors as Nazis? I would conclude that they’re a bunch of failing, desperate nutcases.
Andrew, “the topic of the post” is the application of statistics to temperature measurements. You’re stealing a base when you use “denialist” and “skeptic” (and “alarmists”) while casually including the acronym AGW. It’s a technique favored by Bob-1, and I think is arguably beneath you.
George, actually people who would cling to ideological reasons to deny evidence of AGW used to be called AGW deniers“, “denialist” was coined in a response the “link to Holocaust deniers” charge. To me the term is analogous to people who are in denial of having a drug or gambling addiction, no matter what evidence you put before them, they’re incapable of accepting the reality of the situation. They rationalize that it’s all a big conspiracy, that everyones against them etc, because their addiction to the present situation won’t let them look at things with any objectivity.
Well, thanks for jumping off the deep end there Andrew. How’s the water?
We’re addicted to the present situation, so we can’t look at things objectively? You’re mistaking the main take-away from the UEA correspondence. It’s not the indication of conspiracy, it’s the paranoia. The unmistakable attitude of “We have to protect our phoney baloney jobs here, gentlemen!” They’re the ones who are rationalizing that it’s all a big conspiracy. Big oil. Or something.
As far as clinging to ideology, the subject here is statistical methods and temperatures. “GW”. You’re the one reflexively adding the “A”. Why?
“It’s a technique favored by Bob-1”
What is the technique in question?
I don’t get your point, if you cherry-pick years temperatures above the trend line at the start of a short (~decade long) period and finish with years that have temperatures below the trend line, as denialists often do, you can easily show short term temperature trends well below the overall trend.
Conversely, if you cherry-pick years temperatures below the trend line at the start of a short (~decade long) period, and finish with years that have temperatures above the trend line, as no one does, you can easily show short term temperature increases well above the overall trend line.
I urge you to read the whole thread — it’s not very long. I was critiquing the observations of a pro-AGW blogger named David Kroodsma. Kroodsma performed an analysis through which he claimed to show that global warming was causing nighttime temperatures to rise faster than daytime temperatures. In my critique, I pointed out that (a) he was using a pretty short baseline to come to his conclusion, and (b) trends in min and max temperatures records are highly variable. Along the way I ran across a funny statistical paradox. I pointed out
This is sort of a time series version of Simpson’s paradox. In fact if you study the plot I made, you might notice a curious artifact: for every single pair of consecutive 8-year blocks, for both high and low temperatures, the end of the earlier block’s regression is higher than the beginning of the later block’s regression. I’ll let you ponder the significance of that observation, especially in light of the arguments that the most recent decade still regresses to a positive slope.
And somehow I think you missed the point that these time intervals were not “cherry-picked” to produce this effect. I had 60 years of daily hi/lo data for Memphis handy. I chose to segment it in 8 year blocks so that each block would start on the same day of the year and have exactly the same number of days. And consequently I chose to use the last 56 years — and all of this was done a priori. There was no cherry-picking. For the seven 8-year blocks, there were 4 in which the low temperature rose faster than the high temperature, and 3 in which the high temperature rose faster than the low temperature. So I am skeptical of the significance of Kroodsma’s result.
But nowhere in this critique do I accuse Kroodsma of bad faith or cherry-picking. (That is your invention.) I am merely observing a lack of competence, not goodwill. Like so many climatology researchers, Kroodsma does not bother to establish a null hypothesis or perform a significance test for his results. He simply makes a computation and publishes it with layers of interpretation that are not justified by his data.
if you cherry-pick years temperatures above the trend line at the start of a short (~decade long) period and finish with years that have temperatures below the trend line, as denialists often do,
Speaking of competence, I should point out that your understanding of this statistical phenomenon is less than adequate. It is NOT the case that in the Memphis data the blocks were selected to begin with temperatures on one side of the trend line and end with temperatures on the other side of the trend line. That is NOT what is going on here. If that were a sufficient condition to produce the effect you describe, there would be no way for every pair of blocks to show the effect I described:
Think about it.
OK, when I said I don’t get your point I was trying to square your comment with the topic of the post. I could see that you hadn’t cherry picked the 8 year blocks to engineer short term trends in the Memphis data.
When I said “you can easily show …” I should have said “people can easily show …” I wasn’t clear and my bad.
And you’re right, I don’t claim any expertise in statistics (mind you, people like Briggs can boast about their expertise as statisticians, but if they’re to lazy to actually do the math to support their argument – as Briggs didn’t) their opinion isn’t worth a damn.
Cheers.
(mind you, people like Briggs can boast about their expertise as statisticians, but if they’re to lazy to actually do the math to support their argument – as Briggs didn’t) their opinion isn’t worth a damn.
You know what? You’re right. And I should have commented on this earlier but got sidetracked defending my Memphis data. Although I don’t think Plait did himself any good with his post, I am not too impressed with Briggs’ arguments, either.
One of the treacherous aspects of regression analysis is the subtlety of the different kinds of uncertainty that are associated with the analysis. Strictly speaking, a least-squares regression should take into account the uncertainties attached to each and every data point. Briggs calls this “parameter uncertainty” or “error bars from the estimates” (which is ambiguous at best). I would call it the “input uncertainty” for the regression, or “pointwise uncertainty”.
Now, two points: (a) if the uncertainty is approximately the same for all the data points (say, 0.1 F across the board), then the results of the regression are not changed by including the uncertainty — the pointwise variance can then be pulled out of the chi-squared sum and becomes an overall factor. It is important to account for the “parameter uncertainty” mainly when there are outlying data points with large uncertainties, combined with a large number of better-behaved, better-known values. The other point is (b) for “well-behaved” datasets, i.e. those with approximately uniform uncertainties and a reasonably sound trendline (say R^2 ~ 0.8 or above), it is standard practice to perform the regression without accounting for pointwise uncertainty and let the R^2 value speak to the quality of the regression.
Now, what Briggs is getting at in his post — what he calls “prediction uncertainty” — is the fact that one of the results of regression analysis is a confidence band around the regression line. That is, the regression line as drawn is simply the most likely straight line that explains the data (“most likely” in the sense that it minimizes chi-squared). But there are other regression lines that are almost as good — some above, some below, some with higher slopes, some lower — and these form a probability distribution for the actual trend. The confidence bands around the trendline typical flare out at the lower and upper bounds of the x (date) range. If the trendline is uncertain enough, it is possible that the uncertainties at the highest and lowest dates overlap. That is the possibility Briggs is referring to when he writes, “I don’t know what the prediction uncertainty is for Plait’s picture. Neither does he. I’d be willing to bet it’s large enough so that we can’t tell with certainty greater than 90% whether temperatures in the 1940s were cooler than in the 2000s.”
Now, frankly, if you used other dates, like 1950s vs 2000s, I would bet against Briggs on that one, before running any analysis. But he carefully said “the 1940s” because he knows that global temperatures hit a local maximum in that decade before sliding down to a minimum in the 1970s. So he’s being a little disingenuous here. But likewise, so is Plait, who starts his graph — you guessed it — in the 1970s. If you did a regression from 1940 to 2011, I still think it likely that the 90% confidence band would exclude zero rise. But like any good bettor, I am going to run the numbers before I place my bet.
Everytime I see the term ‘denialist’, I have a sudden urge to order another case of ammunition. It’s more fun than buying indulgences from some modern medeval Pope wannabe.
It’s better than just waiting for the knock at the door.
AS was exactly who I was thinking of.