A couple more questions: Can anyone direct me to any evidence to support a conclusion that atmospheric CO2 concentration is an admissible estimator in the multivariate models of global temperature and that any improved risk functions (such as mean square error or R-squared) resulting from such inclusion are not simply manifestations of Stein’s Paradox? Does the inclusion of CO2 result in any better modeling than would, say, the inclusion of the weights of randomly selected heads of lettuce? Gutman showed in 1982 that Stein’s Paradox is impossible where the data sets are finite, but in this instance time and the other data hardly comprise finite data sets (except possibly in a theological sense). So the admission of totally irrelevant data into a global temperature model could reduce the statistical risk without increasing the information or reliability. You may recall a hilarious article in Scientific American in 1977 [May, I think.] examining the effects on statistical modeling arising from inclusion of Roberto Clemente’s batting average data. As I noted earlier, data from all major global temperature monitoring programs show pronounced and rapid global cooling. Wouldn’t it be a scream if all our efforts to reduce CO2 were actually exacerbating global cooling?