The services looking at climate change use anomalies. Still, compilations of average temperatures for various locations on earth exist– both on line in an a variety of references. The bet resource for you depends on what you want to know. Have you tried weather.com to look up the averages for your zip code? Those sorts of services do give simple weather averages.

They used a one-sided test to show a significant trend term in the unit root model.

Yes. I have avoided one-sided tests for this very reason. Many readers don’t realize that since I use two sided intervals, the tests are supposed to say there is s 2.5% chance a trend lies above the upper interval and a 2.5% chance it lies below the lower interval. (Now, I know you know it doesn’t mean precisely that. Matt Briggs discussed what it means more precisely at his blog. But, wording precisely what it means is difficult, since it really has to do with how often the statistician would be wrong if they applied this test methodology over and over and over etc. Also, it must always be understood that a statistical test result is contingent on the assumptions for the statitical model being correct. But these assumptions are always made.)

Three issues they did not address are:

1. They used a one-sided test to show a significant trend term in the unit root model. A two-sided test with at conventional 1.96 critical value would have made two of the series non-significant. A two sided tests should be used unless there is a very good reason for doing otherwise, as it could be argued that a one-sided tests was used simply as a device to make a conventionally non-significant difference significant.

2. They used a slide-and-eyeball test for showing that “the temperatures recorded in most of the last decade lie above the confidence band of forecasts produced by a model that does not allow for a warming trend.” But in a footnote they admit “The confidence bands in these graphs are calculated using the “dynamic forecast”option in Eviews and they do not incorporate estimation uncertainty.”. A slight increase in uncertainty concerning the parameters and the start point would put all temperatures inside the confidence band.

3. Not using MSU RSS data is a big omission.

Do you have any comment on B & V’s reliance on averages of calendar years rather than monthly data?

I have no idea why they chose to do this. Apparently, they know enough statistics to deal with the autocorrelation, so I would have thought they’d use the available monthly data. Using the monthly data would give more total data points, and also permit them to include the recent downturn in the temperature.

Using the most recent data would, presumably, provide important information. In particular, it would provide more recent data. This is important answer the question “Has the trend changed recently.” As it is, they’ve left out 5% of the data– and in particular, data that are important to the specific question being asked.

That said, I too am surprised Garnaut didn’t ask the economists to consider the questions about the IPCC 2.0C/century projection and/or Rahmstorfs evaluation of the TAR projection. It seems to me, that is the more important question that should be addressed by professional statisticians. I know the questions you asked were related to data comparisons between the AR 4 projections and the TAR projections.

I’ll look at this more on Monday, but right now, those questions were either not understood or ignored by Garnaut and so not addressed by the statisticians he hired. It’s the fourth of July, so I haven’t read the paper in enough detail to say anything more!

Do you have any comment on B & V’s reliance on averages of calendar years rather than monthly data? For 2007, the temperature anomaly in January was between 0.4 and 0.5 C higher than in December, both for the CRUTEM and the GISS. Under a calendar year analysis, the results would have been exactly the same if the progression of GMST anomalies through 2007 had been in the opposite direction. Does this not raise a question about restricting the analysis to annual averages?

I agree. Their approach is fairly similar to yours; it doesn’t address exactly the same questions, but its results don’t contradict yours. I thought the way they went about it was interesting.

Cassanders, as far as I can tell, they’ve included data up to 2007. My reading of their treatment of the recent stuff is that Garnaut asked them the question about the last decade, and they just said they couldn’t find strong evidence of a break in trend in 1998.

Garnaut, who requested the analysis, asked if the positive trend (no specific number) post-2000 was falsified. The report doesn’t say much, but apparently Garnaut says the answer is no.
The text appears to discuss whether or we can falsify a positive trend after 1998.
That doesn’t happen to be the year I look at for a break, because I don’t look for a break. I picked my year based on publication dates of the IPCC documents.

They don’t examine the hypothesis I examined– which is whether or not the data are consistent with the IPCC projection of 2 C/century.

I also don’t falsify a positive trend, and have shown numerous graphs showing the post 2001 data are
a) not inconsistent with the positive trend of about 0.6 C/century since 1900.
b) not inconsistent with the positive trend since 1990.

I haven’t specifically looked at 1976. I’m not surprised there is a statistically break there. The volcanos started erupting in 1963, and 1976 is located in what would commonly be considered the time when the recovery from one of them was expected to start. Then, we start having the cooling aerosols drop out (due to the consequences of the clean air act.)

There is much more scatter and autocorrelation during that period than recently. I think this large autocorrelation is likely a consequence of the volcanic eruptions which introduce a dust veil which persists and drops off.

So, mostly, I don’t see how any of what they say contradicts a single thing I’ve written or found.

It’s just the current trend is not inconsistent with 2C/century or higher. CO falsifies 2 C/century projected by the IPCC.

I’ll have to look at the specific report. But, genearlly, T tests to show whether two trends are consistent tell you whether or not they must be different. But in this case, they could be the same– it’s just the early trend would have to represent a “high” excursion above the underlying rate and the later trend would have to represent a “low” excursion. I may have to do the t-test for two means now. The problem is I need to double check precisely how to do this for trends.

Where BV08 is inadequate is in the test of ANY evidence of change in trend. To assert ‘no significant break’ first you have to establish the test actually has the power to detect a break. There are so few annual data points after 1998 that the power is very low, i.e. it would only detect a very severe downturn. I would have run a structural break test on monthly data from 1979 as see what turns up. With 120 points on each side of a potential break, who knows? When asked for ANY evidence, they should have looked a bit harder, don’t you think!

Cassanders
In Cod we trust

Following up on David Stockwell’s pingback, you might like to look at the time series analysis that he posted. It’s quite readable, even though not all that decisive. The main points I noted were:
1. There is an overall uptrend over the 150 years of data
2. They are strong on identifying a 1976 breakpoint with a stronger uptrend afterwards
3. They identify a ARIMA(0,1,2) model. That’s one step up from your C-O analysis, though I doubt if the reason for preferring second-order is very strong. Anyway, they agree that there is strong autocorrelation.
4. They applied a Jarques-Bera test, and gave no details, but it seemed to pass. Not surprising that they chose J-B, which comes out of their ANU department.
5. Garnaut, who requested the analysis, asked if the positive trend (no specific number) post-2000 was falsified. The report doesn’t say much, but apparently Garnaut says the answer is no.

I think Kenneth over at CA already has a spreadsheet of it for you.

Now I would like to see Lucia include the raw unadjusted data from the CRN1 stations in her statistical tests of the probability of the IPCC forecasts.

My impression is this would involve a huge amount of work on my part. I would need to get the data from each individual station, figure out “the lucia way” of finding bad data. (There will always be bad data at even the best station.) Then, I would need to figure out how to weight each station etc.

This is the sort of thing that NOAA/NCDC, GISS and HadCrut assign to teams of people. The project may be do-able, but I’m afraid creating a “lucia” data based based on the stations Anthony’s volunteers have found to be good is more than I’m willing to take on.

But… if someone were to form a group blog etc. to do it, that would be great.

If the US isn’t warming, no one is. We’re number one dammit!

I think the IPCC forecasts will be mathmatically improbable. Also, it may be that there has been no material recent warming in the US. The most accurately suveyed area in the world.”>

http://www.climateaudit.org/?p=3169#comment-266890

You just better hope they don’t put a special carbon tax on cookies!

There was a problem with my Vostok questions. I had assumed that a dataset I hadn’t seen was as robust as one I did see. This explains why I had never seen the question. There simply isn’t enough CO2 data for the recent (20000 years) past. Had I bothered to look instead of assuming, I would have went, “Oh, well that won’t work.” *shrugs*

The other questions– I know there are nothing wrong with those questions. It is just the way my mind works when it comes to “threats”. My sister tells me I’ve been brainwashed to think that way. Brainwashing, conditioning, and training are all pretty much identical concepts. The differences involve intent and technique. Technically, I was trained to view threats a certain way. So when they occur, little thought is required to take action.

I recognize that changes in climate can be a threat to national security. For an example, a potential of a modern European “little ice age”, prompted the creation of a plan for a military response to the destabilization caused by the event. I would rather spend 500 billion dollars to prevent the event which was quite likely to occur, than to spend 500 billion plus loss of life responding to the destablization. But, I wouldn’t want to spend a dime, if there wasn’t a reasonable chance of neutralizing the threat, or if the threat wasn’t quite likely.

So, do you have any opinions on the questions?