Fun with correlations!

We are forever being bombarded with apparently incredible correlations of various solar indices and climate. A number of them came up in the excoriable TGGWS mockumentary last month where they were mysteriously ‘improved’ in a number of underhand ways. But even without those improvements (which variously involved changing the axes, drawing in non-existent data, taking out data that would contradict the point etc.), the as-published correlations were superficially quite impressive. Why then are we not impressed?

To give you an idea, I’m going to go through the motions of constructing a new theory of political change using techniques that have been pioneered by a small subset of solar-climate researchers (references will of course be given). And to make it even more relevant, I’m going to take as my starting point research that Richard Lindzen has highlighted on his office door for many years:

That’s right. Forget the economy or the war(s), the fortunes of the Republican party in the US Senate are instead tied closely to the sunspot cycle.

“Oh yes”, the sceptics might say “but that’s just a couple of cycles and doesn’t use up-to-date numbers. What happens after 1986?”

Well, that is a little problematic, however, the good early correlation is obviously still important (r=0.52! 1960-1986) and so we should be able to refer to it over and over again without noting that it breaks down subsequently (cf. Svensmark, 2007 referring to Marsh and Svensmark (2000)). But more importantly, it just demonstrates that the theory needs a little adjustment.

Let’s look at the second half of the record. Well, there’s another strong correlation for that period as well (r=-0.63, 1988-2006). Only this time the correlation is inverted, but that shouldn’t be surprising to anyone – solar-senator effects are complicated!

If we now put it all together, we can see that there is a reasonable match over the whole period…. well, except that break in the period 1984 and 1988 and, unfortunately, last year’s elections didn’t fit the pattern either. But 1984-1988 was Ronald Reagan’s second term and clearly no theory of Republican senators can ignore that. We therefore propose that the ‘Ronald Reagan second term phase shift’ combined with the change of sign of the Hale solar magnetic cycle in 1986, obviously changed the dynamics. This kind of phase shift is frequently seen in solar studies (cf. Landscheidt and many others), where it is rarely taken as a sign that two time series with decadal spectral power are in fact completely independent. Finally, it is permissible to leave off the more recent data points (cf. TGGWS) for “graphical convenience”. So after just a little work, we have managed to rescue the original theory to match a much longer amount of data:

Some readers may scoff and suggest that in the absence of any mechanism, these powerful correlations are numerological artifacts arrived at using post hoc fallacious reasoning that have no predictive capability. That might appear to be a valid argument. However the ultimate test will of course be experimental. On the basis of these intriguing results, we propose exposing Republican senators to varying levels of cosmic rays in a basement and monitoring their electability. Any refusal by the funding agencies or ethical review panels to support this would simply be confirmation that the political science establishment are scared of what this research would imply for their so-called “consensus”.

Convincing, eh?

The data for sunspots and senators can, I’m sure, be manipulated even more effectively than I’ve done here. I’ve made no use of various lags or filters (which can be altered as you go along cf. Friis-Christensen and Lassen (1991)), or of partial detrending (cf. Marsh and Svensmark (2003)), or of splicing of unconnected data sets (cf. Svensmark and Friis-Christensen 1997, Nir Shaviv). More ideas could be taken from “New evidence for the Theory of the Stork” (Höfer et al, 2004)”. A special RealClimate commendation for anyone who can do better!