# Global Warming statistical analysis

The concept of anthropogenic global warming comes with massive policy implications attached. My gut reaction is that it’s grossly exaggerated at best. But it would be unwise to discount the theory without examining it thoroughly. While my knowledge of science is very limited, I would like to look at it through statistical methods, which would require the following:

I. Data
a. Global temperature record; methodology and uniformity of data collection locations over time in obtaining this data.
b. CO2 record; same as with temperature.
II. Analysis
a. Run a regression with a linear and a quadratic term. Confirmation requires a significant positive coefficient for the x term and the absence of a significant negative coefficient for the x^2 term.
b. Introduce quantifiable non-anthropogenic factors into the regression. Check for multicollinearity.
c. Ignorance-fighting! Profit! Beer!

I’m set on the number-crunching aspect of it via access to Stata at school for another month. Finding the complete data presents more difficulty. Anyone with me?

Nope, not with you. Do you have any theory that supports your choice of a simple linear regression? The “true model” is out there somewhere, but if you don’t have the understanding to model the data generation process, how can you presume to test a hypothesis? The data are time series: I imagine you have not even considered serial correlation or heteroscedasticity.

I am going to go out on a limb here and assume that you want to run a regression because it is the only model you have understand and can implement in STATA.

The IPCC report is essentially exactly what you are proposing. But note that it had several thousand people going over the results of tens of thousands of studies and merging them to produce best guess numbers from the relevant studies.

Doing it by yourself is rather meaningless and moreover would take several lifetimes. And individually picking and choosing studies to minimize the total amount of data you have to process has no way of leading to a better result than what you can get from the official reports.

If you doubt the UN results, you can feel free to sample the EPA or the Federal Climate Change group’s reports to congress. But even those pretty well admit that the UN has achieved the current best review of all current information in the IPCC report.

There are thousands of climate scientists working on the problem of understanding AGW and their techniques have advanced well beyond the simple statistical ones that you propose here. In fact, there are various studies using climate models that have attempted to back out the climate’s sensitivity to CO2 by looking at the instrumental temperature record. Unfortunately, however, the constraints one gets on the climate sensitivity in this way aren’t that strong, mainly because there is too much uncertainty regarding other forcings, in particular, that due to man-made aerosols. Better constraints on climate sensitivity come from other methods, such as looking at the climate response to the eruption of Mt. Pinatubo or looking at various paleoclimate events, in particular the difference in temperature and forcings between the last ice age (“last glacial maximum [LGM]”) and now.

From the theoretical side of things, one can get estimates by putting all the physics that we understand into climate models and seeing what climate sensitivities or range of climate sensitivities they come up with (and, particularly, what range also produces the best agreement with present day climate). As one example, the climateprediction.net experiment varied various parameters in a climate model within what experts estimated to be plausible ranges and found that they could not get a climate sensitivity below about 2 C. (They could get ones as high as 11 C, although these were in the tail of the distribution and others have pointed out that there was no attempt to look at these individual runs to see if the general present-day climate they produced was in good agreement with observations.)

Combining all of these ways of estimating the climate sensitivity is what has led the IPCC to conclude that the value for equilibrium climate sensitivity (i.e., change in global temperature under a doubling of CO2 levels) is likely (estimated 66-90% chance) to be in the range of 2 to 4.5 C and very unlikely (<10% chance) to be less than 1.5 C.