How to cherry pick data
Here’s the banner image for the Global Warming Policy Foundation, Nigel Lawson’s ‘think’ tank.
On the left of the image, that’s a cute data plot of 21st Century global mean temperature (2001-2008), and, plainly, it shows that the temperature is in decline over the seven year period.
But why only seven years of data, and not 130 years of data? For the forward looking chaps at the GWPF, one supposes that would be just too last millennium. It’s 2010, after all. Why look backward?
Here’s why not:
The Global Warming Policy Foundation got it right with their banner image; that is how to cherry pick data. Don’t show, don’t even hint at, the existence of something that contradicts your message, like 130 years of temperature data with a really clear positive trend. Just ignore it or deny it.
How not to cherry pick data
With that in mind, take a look at this really quite baffling graph from the website of the Orwellian-named Canadian ‘Friends of Science’:
This series runs from 1979 to 2009. It shows data from two averaged satellite temperature records, with the CO2 concentration in green. Both birds have only been flying since the end of the 1970s, hence the data series starts in 1979. “Surface temperature data is contaminated by the effects of urban development”, the ‘Friends’ tell us, which one supposes must suffice as explanation for their only using satellite data. As an aside, one wonders if science really needs such friends as these, able to dismiss 130 years of surface temperature measurements and scientific endeavour with an imperious wave of the hand. But let us persist, and see what the satellite data shows; perhaps a best fit over the 30 years of satellite measurements will reveal some interesting information.
Well, the ‘Friends of Science’ were good enough to plot a best fit (that odd little purple line on the right), but one that started on January 2002, not 1979:
“The best fit line from January 2002 indicates a declining trend.”
Indeed it does, dear ‘Friends’, but why a best fit for that interval? 1998 to 2009 would have really made your point. Plotting a best fit from 2002, while plainly ignoring the rest of the data, just looks silly.
Where the ‘Friends of Science’ went wrong, clearly, was in revealing the existence of a full 30 years of satellite temperature measurements. If their x-axis had started in 2002, it would have made for a much more subtle cherry pick than displaying 30 years of data and adding an orphaned best fit line that doesn’t cover even a third of the whole plot.
I don’t know, it’s as if the ‘Friends of Science’ aren’t really committed to denialism.
Take a tip from the GWPF’s playbook: if you are going to cherry pick data, don’t blatantly show the audience that is what you have done, and don’t even hint at the disconcerting existence of such findings as this:
Shhh…
