Statistical Analysis
http://scholar.lib.vt.edu/theses/available/etd-11292004-134819/unrestricted/bodymatter_akobundu.pdf
This document analyses two common data sets as used by used by Hansen and Jones and looks at Linear Regression tests and Null Hypothesis Confirmation.
If finds that the series are not Linear and not correctly specified and that the causation is not establsihed, requiring more research as the the real drivers of the variations. ie NOT CO2.
Page 33
Section 4: Conclusion
To answer the question posed at the outset, the normal linear regression model can assess the evidence for a warming trend when the model is correctly specified. The dual goal of this paper was to illustrate why the static model may be invalid for some global temperature series and to propose an alternative formulation that is statistically adequate for the data.
Concerning the first goal, the results indicate that the static model is misspecified for two widely-used temperature series and is not reliable for drawing conclusions about trends in surface temperatures using these series.
The model provides estimates of a statistically significant warming trend in temperature series that have been widely 34 reported in international policy arenas. The results indicate that these estimates may overestimate or mischaracterize the trend in temperature anomalies.
Concerning the second goal, the results indicate that the temperature series exhibit nonlinear trends rather than linear trends.
A nonlinear trend with an overall positive slope provides a different picture of global warming compared to a linear trend with a positive slope.
Given the recent sharp upward trending temperatures predicted by the dynamic model, some may claim that the better specified model provides more impressive and urgent evidence of global warming than the linear model.
The model results neither support nor refute such a conclusion because, while the model provides statistical evidence of nonlinear trends in the data, it does not explain the cause of the trends nor indicate whether the trends are temporary or permanent.
In a sense, the trends in the statistical models are a measure of the researchers’ ignorance about certain characteristics of the data generating process.
It is up to climate scientists to explore alternative possible theories, as well as tests of these theories, that can explain the confirmed nonlinear patterns in the data.
This is the basis of John Christy's comments in the reported article that our ignorance is simply enormous and biased to one direction and we simply don't understand the drivers.