My own version of the same idea:
Forecasting models are tested by "back forecasting". You make your model based on historical data, and test it by covering up the last few/several periods. Then you uncover those last periods and see how close you came.
The difference between predicted and actual is the error in your model. You can take both the mean and the standard deviation of the error. Ideally, the mean is 0 and the standard deviation is small.
Based on the standard deviation, you get a good idea of how close to "spot on" your forecast is, because you can statistically set limits on how much error is "normal".
What has actually happened is that the forecasts for recent years have fallen outside the limits to normal random variation. They are far enough out and there are enough of them that the probability of getting so much error just by normal random variation is just about nil.
And that's all you can do with statistics. You never prove anything. You just choose to believe one thing or another depending on probability.
Super short version: The warmer climate model is severely busted and not of much use in predicting what will happen next.
It doesn't matter how beautiful your theory is, it doesn't matter how smart you are. If it doesn't agree with experiment, it's wrong.
Richard P. Feynman
Last edited by denton; 02/12/16.
