I haven’t checked out Iain Dale’s CV, but I suspect it does not include a period studying statistics. Because he seems to believe in a recent post that taking one month every ten years from one location in the UK is a sufficient reason to justify statements about the effect of carbon dioxide throughout the entire globe over a 150 year period. Yup. Coming to a dinner party near Iain, some poor soul will be bored by the story of how we’re not warming up, because 1868 was a warm June in Oxford.
One useful consequence of this bunch of nonsense is that it led me to the Met’s collection of data. The first thing that any statistician will notice is variability. Taking one region at random (Kirkwall) and a month (May), the standard deviation over the period in question was 0.95 centigrade. Given that a 2 degree warming is what is bothering people, that will happen between year 1 and year 2 around 3% of the time, anyway. So you get a lot of real scattering.
For example, consider Stornoway. Just taking its annual average instead of one month reduces the scattering:
Now, what about the whole UK? One of the problems is that weather stations have been added over the years. And they seem to have been added more in the south, so that just adding them raises the average. In theory, I could take this into account in about 20 minutes – just adjust the averages by how much Guernsey is warmer than Glasgow, say. But since this is aimed at a non-statistician like Dale, such activity would raise suspicions. So I just take the 7 stations that have recorded data from 1890 to 2008, which are:
and I average over the whole year for all seven. So, each year represents 84 data points, not 1. In theory, this should reduce the standard deviation by a lot – but since there is a fair amount of correlation between different parts of the UK and different months in the same region, not by very much. StDev falls to 0.45. This is what it then looks like:
Note the equations. The top one indicates that the slope of the line is about 0.7 C per century. However, if you took the last 30 years of so, and did THAT line, you would get a steeper rise:
Plain old extrapolation is a lousy, crude instrument. It also only explains a portion of the movement. El Nino and all that. The R-squared figures indicate that just 20-40% of the annual change in temperature is accounted for by that small (0.01-0.02) degree change in the average – there is a lot of Gulf Stream in there. The standard deviation each year is about 20 times as big as the creeping-up average. This is why some idiot remembering how it was warmer a few years back is such an idiot. It is rather like someone refuting the proposition “Man U are better than Leyton Orient” with a single attempt on goal in an FA Cup 3rd round game.
I have spent all of one hour on this. I am good with figures, and still think I would need many more hours with the global dataset to make definite claims from it – there are many pitfalls, and possibilities of skewed selection, in proving things statistically. However, the very ease with which blundering loudmouths think they can *prove* from single data points, that there is no global warming, again proves to me that the choice of what to believe is not about the science or the stats.
It becomes increasingly clear: sceptics do not believe what they believe because of access to a better data set, an improved set of scientific techniques for treating that data, a better theory for explaining it, or indeed better ethics. They believe it because the vastness of the Internet, and the intrinsic need for human beings to find results that confirm their own baises. They believe it because they want to – and my post, full of graphs, is unlikely to show them otherwise.
UPDATE: Unity on Liberal Conspiracy has done an even more convincing job here