That’s a good overview. Especially the last third. I now have somewhere to point my numerate friends other than to just say, “Go read the last two years of Climate Audit.”
Yeah, its funny that this sort of analysis passed off as cutting edge just 10 or 15 years ago.
What I find ironic is that something so straightforward has so many pitfalls of interpretation and opportunities to place a thumb on the scale.
Similarly, someone (I think it was Lucia at The Blackboard) had a neat post with a meta-chart showing the results with a simple statistical test of whether the IPCC forecast hypothesis can be rejected, varied by choice of start year. Equally reasonably sounding choices of start year produce different results.
I wish I’d seen this example during my intro to econometrics lecture on the dangers of data mining. When you’d in the middle of an analysis, it’s so hard to distinguish between adjusting your procedure to accommodate an unforeseen problem with the data and tainting the results.
Yeah. The only way out of this mess is to have out of sample tests. The proxies end in 1980 for some reason… maybe someone should go collect some more tree ring data and see if forecasts using Mann’s 1998 estimated coefficients do a good job matching the instrument data for the last 30 years.