Two comments from UC on smoothing CET using Mannian smoothing, a technique peer reviewed by real climate scientists (though not statisticians).
I think these coldish years do matter, maybe now there will be some advance in smoothing methods. mike writes (1062784268.txt) ( I think this is somewhat related to CET smoothing (?) )
The second, which he calls “reflecting the data across the endpoints”, is the constraint I have been employing which, again, is mathematically equivalent to insuring a point of inflection at the boundary. This is the preferable constraint for non-stationary mean processes, and we are, I assert, on very solid ground (preferable ground in fact) in employing this boundary constraint for series with trends…
I assert that a preferable alternative, when there is a trend in the series extending through the boundary is to reflect both about the time axis and the amplitude axis (where the reflection is with respect to the y value of the final data point). This insures a point of inflection to the smooth at the boundary, and is essentially what the method I’m employing does (I simply reflect the trend but not the variability about the trend–they are almost the same)…
And now this leads to following figure:
Jones also mentions CET:
Normal people in the UK think the weather is cold and the summer is lousy, but the CET is on course for another very warm year. Warmth in spring doesn’t seem to count in most people’s minds when it comes to warming.
And later here:
Yes, extrapolations are problematic if someone bothers to check those later:
Can’t understand what Mann means by ‘preferable constraint for non-stationary mean
processes’.. I’d prefer no smoothing at all if there is no statistical model for the process itself, something like this maybe:
Update (UC, 8 Jan 2011)
Code in here .
For CA readers it is clear why Minimum Roughness acts this way, see for example RomanM’s comment in here (some figures are missing there, will try to update). But to me it seems that the methods used in climate science evolve whenever temperatures turn down (Rahmstorf example is here somewhere, and you can ask JeanS what has happened in Finnish mean temperature smooths lately).