UC writes in that there’s another Mannian problem:
And GRL08 ( http://www.meteo.psu.edu/~mann/shared/articles/MannGRL08.pdf ) is the prologue. Tried this new smoother ( http://www.meteo.psu.edu/~mann/smoothing08/lowpassadaptive.m ) with HadCRUT monthly and f=0.0104,
[smoothedbest,w0,w1,w2,msebest] = lowpassadaptive(HadM, 0.0104);
got this figure,
But the weights are [w0 w1 w2]
0.4100 0 0.5500
Surprisingly, this does not agree with the text, where constraint sum(w_j)=1 is specified. I guess there’s something wrong with Mann’s exhaustive search,
I hope he will fix this soon.
UPDATE Sept 3, 8 am: It didn’t take long. As noted in a comment below, Mann changed the online code within hours of this being posted at CA to the following (note the difference in the third line). There is no annotation of the change nor credit to CA.