Skeptics are not statisticians: 1% of a data set does not disprove the whole data set.

A friend of mine likes to provoke me by sending links to posts and articles that contradict the established science on climate change. Recently he pointed me to this post that appeared in Watt's Up With That: NCDC data shows that the contiguous USA has not warmed in the past decade, summers are cooler, winters are getting colder.  My friend accompanied this link with this comment:
Well, the figures are in from the NOAA National Climate Data Center and the rest is just high school math. While US CO2 emission has gone up precipitously in the last 100 years the average temperature in the US has gone down. Of course you could look at just the last 15 years and ignore the rest. In this case CO2 has still gone up a lot and temperature was flat. Does this fit with saying the controversy is over and global warming is an established fact.

I countered with some points, but only later realized the biggest flaw with this reasoning, which I'll get to shortly.



The WUWT post does not outright dispute global warming, but rather presents these data as
somehow bringing into question the validity of global warming through implication more than direct assertion.  That's the way with many of the more careful skeptical bloggers: be careful not to state outright what we know to be false (e.g., global warming is not happening). Rather throw up some graphs and spend time nitpicking at details to create the illusion that there's a problem with the overall argument.  In this post, there is a subtle suggestion that the existence of a counter-trend in data somehow calls into question the conclusions of the BEST report

But that's an absurd position.  Keep in mind that all temperature data worldwide are included in the BEST data.  So the data upon which this post is based are a subset of the BEST data.  Essentially for this post, they ignore 99% of the BEST data, find a selected set that shows different results than the full set, and subtly imply that somehow the tiny subset calls into question the entire set.  Since when does a subset supercede a full set of data?  Yet that's the implication of this post.

As an analogy, say polling is done and 60% of men prefer women with green eyes.  Then someone looks at just the data of left-handed men.  Only 25% of them prefer women with green eyes.  That may be interesting to note, but it does not have any bearing whatsoever on the fact that the majority of men prefer women with green eyes.  It's still 60%.

Likewise with BEST.  The full data agree with the other well-known temperature data sets that show an overall warming of about 1 degree C over the last century.  This particular subset may tell us something interesting about a tiny portion of the globe during a particular period in time, but it does not change the conclusion of the overall data one miniscule iota.

"Does this fit with saying the controversy is over and global warming is an established fact?" my friend asks.  That's a senseless question.  The BEST data strongly support the established fact of climate change.  A tiny subset of that data changes nothing.

No comments:

Post a Comment