Terry Oldberg says:
December 18, 2011 at 1:13 pm
“In their paper, Foster and Rahmstorf reveal either a lack of understanding of or a distain for the scientific method of inquiry by presenting models that are neither statistically validated nor susceptible to statistical validation. Models of this kind lie outside science.”
F&R are right. You should not apply statistics to computer model runs that are just behaving as they are programmed to behave. It is hard to calculate confidence intervals based on confidence intervals of the observations of the input data and parameters. In theory you could try with Monte Carlo methods and running large number of runs, but you still have large aggregated computing errors due to lack of precision.