View Single Post
nesp
#22
Jan24-07, 04:00 PM
P: 59
<Guy Callendar, who wanted to proof greenhouse effect, cherry picked those close to the desired hockeystick and ignored dozens of others with a completely different story>

Yes I've seen that, and noted that the error was embedded in his biased normalization of data for principal component analysis. I like Karner's approach using ARIMA models from the link you provided, not that they are necessarily better than PCA but time domain methods are more transparent and harder to cherry pick.