Friday, June 20, 2014

Matt Rogers Can't Math

Another fake climate skeptic has published a misleading article, disappointingly this time in the Washington Post.  Matt Rogers says that there has been a deceleration in surface warming.  To show this, he gives the first difference graph of both surface temperature data from NASA (GISTEMP) and NOAA (NCDC).

His claim is false.  The standard errors of the trends in the first difference graphs are each greater than their respective trends, meaning you can't say even at 1-sigma that the trends are significantly different than zero (and you most certainly may not make a claim at 2-sigma).  For NOAA's data, the trend and standard error are -0.0043 and 0.0055 respectively; for NASA's data, the trend and standard error are -0.0055 and 0.0068 respectively.
NCDC: -0.0043±0.0110 (units ~˚C/year^2)
GISTEMP: -0.0055±0.0137

It is also unclear what data he is using for the GISTEMP dataset.  The 2001-2000 value should not be as low in his graph as it is; the one above is correct.  It's interesting that even with that correction, the trend is still very insignificant.

Rogers also tries to shield himself from criticism of cherry picking, by saying that he could have picked 1998 as a start year.  1998 was a very warm year, and Rogers thinks that this would have amplified the trend line.  As a fake skeptic, it is par for the course he would try to protect against accusations of cherry picking that dreaded year.

This is also a false claim anyway, and embarrassingly so.  If you start with a warm year like 1998 (contrasted to a cold year like 2000), and then do first differencing, you're going to start with a very low datapoint.  The result is not an amplified negative trend, but in fact a more positive trend.  Anyone that had actually graphed out the data would know that.  Anyone that can do basic math would know that, in fact.
NCDC: 0.0030±0.010
GISTEMP: 0.0023±0.0126
These are also not statistically significant.  In fact no first-difference trend is statistically significant, for at least a couple decades back.

Rogers doesn't know what he is talking about.  And this is a rather funny way of illustrating that fake skeptics in general don't know how to handle 1998.

[Edit: The extra attention this post seems to be getting has encouraged me to quickly add trend lines to these graphs.  Hopefully that helps make things a bit clearer!  I've also added indented "tables."  Thanks to Tom Di Liberto for the plug.]

No comments:

Post a Comment