Comments on: Spurious Regression Random Walk http://landshape.org/enm/spurious-regression-random-walk/ The power of numeracy Mon, 15 Mar 2010 09:48:22 -0500 http://wordpress.org/?v=2.8.6 hourly 1 By: Tim Curtin http://landshape.org/enm/spurious-regression-random-walk/comment-page-1/#comment-146859 Tim Curtin Thu, 17 Jul 2008 04:54:49 +0000 http://landshape.org/enm/spurious-regression-random-walk/#comment-146859 David: the global temperature sets prescribed to Breusch and Vahid by Garnaut begin from 1850 or 1880 and deliberately did not include the satellite record which is now 30 years. The useful attribute of the 19th century years from 1850 to after 1900 is that there were no weather stations in tropical Africa or even central America and much if not most of tropical Australia. This conveniently makes those years seem globally "cool", just as losing all those Siberian stations makes the NH seems warmer than it was after 1990, as the satellite measurements also show. Using post 2002 GISS data also introduces bias from 1980 to 2008 after all Hansen's rewriting of climate history to make the 1980s look cooler and the 1990s hotter. David: the global temperature sets prescribed to Breusch and Vahid by Garnaut begin from 1850 or 1880 and deliberately did not include the satellite record which is now 30 years. The useful attribute of the 19th century years from 1850 to after 1900 is that there were no weather stations in tropical Africa or even central America and much if not most of tropical Australia. This conveniently makes those years seem globally “cool”, just as losing all those Siberian stations makes the NH seems warmer than it was after 1990, as the satellite measurements also show. Using post 2002 GISS data also introduces bias from 1980 to 2008 after all Hansen’s rewriting of climate history to make the 1980s look cooler and the 1990s hotter.

]]> By: admin http://landshape.org/enm/spurious-regression-random-walk/comment-page-1/#comment-142191 admin Sun, 13 Jul 2008 00:44:17 +0000 http://landshape.org/enm/spurious-regression-random-walk/#comment-142191 Bob, I have shown that 2 of the the more conventional two-sided tests would not have been significant. That is among the facts. Bob, I have shown that 2 of the the more conventional two-sided tests would not have been significant. That is among the facts.

]]> By: Bob http://landshape.org/enm/spurious-regression-random-walk/comment-page-1/#comment-142186 Bob Sun, 13 Jul 2008 00:28:16 +0000 http://landshape.org/enm/spurious-regression-random-walk/#comment-142186 There are no facts, only skepticism. There are no facts, only skepticism.

]]> By: UC http://landshape.org/enm/spurious-regression-random-walk/comment-page-1/#comment-140470 UC Thu, 10 Jul 2008 16:51:36 +0000 http://landshape.org/enm/spurious-regression-random-walk/#comment-140470 Breusch and Vahid does not cite Shumway and Stoffer, "Time Series Analysis and Its Applications" (2006) p. 62, "In this case it appears that the differenced process may be white noise, which implies that the global temperature series is a random walk" or Mizon, "A simple message for autocorrelation correctors: Don't", Journal of Econometrics (1995) 69 ..."'autocorrelation correction' is invalid in general, and cannot be justified on the grounds of 'robustifying' estimation against the presence of residual serial correlation" Breusch and Vahid does not cite

Shumway and Stoffer, “Time Series Analysis and Its Applications” (2006) p. 62,

“In this case it appears that the differenced process may be white noise, which implies that the global temperature series is a random walk”

or

Mizon, “A simple message for autocorrelation correctors: Don’t”, Journal of Econometrics (1995) 69

…”‘autocorrelation correction’ is invalid in general, and cannot be justified on the grounds of ‘robustifying’ estimation against the presence of residual serial correlation”

]]> By: Truthman http://landshape.org/enm/spurious-regression-random-walk/comment-page-1/#comment-137936 Truthman Sun, 06 Jul 2008 23:25:12 +0000 http://landshape.org/enm/spurious-regression-random-walk/#comment-137936 The trouble is when you have people claiming disaster in100yrs they dont deal in facts,the reality is n ne can forcast that far ahead accurately nor can they ever change nature or the weather .a very good article !you know its a scam when nothing is based on fact . The trouble is when you have people claiming disaster in100yrs they dont deal in facts,the reality is n ne can forcast that far ahead accurately nor can they ever change nature or the weather .a very good article !you know its a scam when nothing is based on fact .

]]> By: david http://landshape.org/enm/spurious-regression-random-walk/comment-page-1/#comment-137660 david Sun, 06 Jul 2008 20:20:05 +0000 http://landshape.org/enm/spurious-regression-random-walk/#comment-137660 Admin, If you take a temperature in Alaska and one in central Australia, and average the two, what do you get? World average temperature? Of course not. If you include more stations in the average it just gets worse, not better. How do you know which stations to include? For example, how about the contribution of Siberian stations since the collapse of the Soviet Union? And thousands of similar questions. The central limit theorem is no help here: as you say an average from *independent* populations would lead to a stationary, normally distributed variable. But only a statistician could believe for a moment that these are independent. The Northern Hemisphere tends to have winter all at the same time, which is not the same as the Southern Hemisphere. What are the odds on that for independent variables? The gulf stream regularly takes warm water from the tropics to North Western Europe. Check out the population density at similar latitudes in Canada. And so on, again with thousands of interdependencies. Can I specify all of those? No, if I could I would have a reliable method of long-term weather and climate prediction. I think everyone knows that weather prediction is unreliable beyond a few days, so I think it is safe to say that no-one has such a model. To put it another way, if local climate data are in fact independent observations then we have no global anything. Just a bunch of local climates. Admin,
If you take a temperature in Alaska and one in central Australia, and average the two, what do you get? World average temperature? Of course not. If you include more stations in the average it just gets worse, not better. How do you know which stations to include? For example, how about the contribution of Siberian stations since the collapse of the Soviet Union? And thousands of similar questions.

The central limit theorem is no help here: as you say an average from *independent* populations would lead to a stationary, normally distributed variable. But only a statistician could believe for a moment that these are independent. The Northern Hemisphere tends to have winter all at the same time, which is not the same as the Southern Hemisphere. What are the odds on that for independent variables? The gulf stream regularly takes warm water from the tropics to North Western Europe. Check out the population density at similar latitudes in Canada. And so on, again with thousands of interdependencies. Can I specify all of those? No, if I could I would have a reliable method of long-term weather and climate prediction. I think everyone knows that weather prediction is unreliable beyond a few days, so I think it is safe to say that no-one has such a model.

To put it another way, if local climate data are in fact independent observations then we have no global anything. Just a bunch of local climates.

]]> By: admin http://landshape.org/enm/spurious-regression-random-walk/comment-page-1/#comment-137261 admin Sun, 06 Jul 2008 05:15:37 +0000 http://landshape.org/enm/spurious-regression-random-walk/#comment-137261 Carl, It would help your argument to explain why temperature from independent populations would lead to non-stationary random walk variable. By the central limit theorem, an average from independent populations would lead to a stationary, normally distributed variable. Carl,
It would help your argument to explain why temperature from independent populations would lead to non-stationary random walk variable. By the central limit theorem, an average from independent populations would lead to a stationary, normally distributed variable.

]]> By: Carl W http://landshape.org/enm/spurious-regression-random-walk/comment-page-1/#comment-137238 Carl W Sun, 06 Jul 2008 02:37:55 +0000 http://landshape.org/enm/spurious-regression-random-walk/#comment-137238 Once again, this is why average global temperature is irrelevant. By looking at global temperature, we're merely combining a whole set of diverse regional trends that need to be looked at independently. Ocean heat content, regional surface temperature trends, and regional atmospheric temperature trends are what matter - not a meaningless global average surface temperature. See also: http://www.uoguelph.ca/~rmckitri/research/globaltemp/globaltemp.html Once again, this is why average global temperature is irrelevant. By looking at global temperature, we’re merely combining a whole set of diverse regional trends that need to be looked at independently. Ocean heat content, regional surface temperature trends, and regional atmospheric temperature trends are what matter – not a meaningless global average surface temperature.
See also: http://www.uoguelph.ca/~rmckitri/research/globaltemp/globaltemp.html

]]>