from: drdendroatXYZxyzo.columbia.edu (Edward R. Cook)
subject: Re: Your paper
to: Keith Briffa <k.briffaatXYZxyz.ac.uk>
One would think that your paper would be a slamdunk. But we all know how
incredibly fickle Nature is. I got a call from Philip Newton (no doubt
Isaac's direct lineal descendent. Nature wouldn't have it any other way!).
He asked me to review a 2-page comment-style submission on Bayesian
regression methods applied to tree-ring analysis. I wonder if it is from
Van Deusen. He is the only one who has ever played with that stuff with
tree rings. Seems the tone of the comment is that tree-ring scientists have
been getting it wrong for all these years by using standard OLS. I'll let
you know what it says when I get it.
Ken is working on a general non-linear regression routine for fitting
non-linear growth models to tree-ring series. He is generalizing the
Hugershoff and general exponential models to allow for a positive
asymptote, like the modified neg expon curve, and for allowing for an
offset from the origin. As it stands now, the Hugerhoff and gen expon
models assume that the first ring is formed at year t=1. We know that this
is not the case in almost all cases. How it will all work in practice is
yet to be seen. I also found an error in the modified negative exponential
curve routine in the Tucson DPL version of ARSTAN. Seems it would not
iterate properly and kicked out too quickly often giving near-linear fits
to exponential growth trends. I validated it, with the help of Dave Meko,
with 6 independent programs like MATLAB, SPSS, and Kaleidagraph. So, I sent
my Fortran code to Richard and the error has been corrected. The corrected
version of DPL ARSTAN is now on the Tucson ftp site. If you or anyone else
uses it, please tell them. Henri suggested that I put out a notice on the
forum, but I reckon that that is something that Tucson should do. Afterall,
it's their baby ... with a few birth defects.
I have been able to change my schedule and flight for Australia, so now I
can make the session. I haven't hear from Phil yet on this, but I guess he
will be pleased.