Sunday, December 6, 2009
I read the Wegman Report last night, and I think I now understand how Michael Mann's statistical methods produce the Hockey Stick from random input data. I have described the process here. In brief: his method assumes the accuracy of the global surface trend (CRU trend) and calibrates all other series with respect to the rising temperatures of the 20th century. If a series shows a rise in the 20th Century, it is added to the result with greater weight. If a series shows a decline in the 20th century, the decline is flipped around so that it becomes a rise. Data with no trend in the 20th century is added last and with less weight. The result is a hockey stick almost all the time when you combine ten random (red nose) series with the CRU trend.