Heller and the 5 Year Glitch

A quick lesson in how to deceive with statistics, from Tony Heller. In the post The Pause Is Real. NASA Temperatures Aren’t., Steve Goddard (aka Tony Heller) says

Satellites show no warming for twenty years – but fake NASA surface temperatures show lots of warming.

With this graph from Wood For Trees to illustrate the point.

Tony Heller's graph showing no warming in UAH 6 data, since 1997.

It's true that NASA's data shows lots of warming, but is it really true that satellites show no warming. Lets check this using the same data.

Graph showing some warming in UAH 6 data, since 1997. Wood For Trees

UAH definitely shows some warming, not the flat line Heller's graph shows. Both graphs use the same data, both show the temperatures using a 5 year rolling average, but whereas Heller calculates the trend line using the averaged data, I calculate it correctly - using the monthly data.

Using smoothed data is not the way to calculate a regression. In particular it reduces the influence of the start and end points of the period. In this case the first and last 2.5 years have less importance in determining the trend line than the middle 15 years, with the influence reducing as you move close to the start and finish. The most recent 30 months were quite a bit warmer than the first 30 months, so reducing their importance is sufficient to turn an overall warming trend into a slight negative trend.

This explains why the true linear regression looks out of place when compared with the 60 month rolling average. If we look 12 month rolling averages the picture becomes clearer.

Wood For Trees

To put some figures on the above, since the start if 1997, UAH 6 has been warming at the rate of 0.64 °C / century, compared to GISS's rate of 1.72 °C / century. The old UAH 5.6 has been warming at the rate of 1.60 °C / century.