Servien remi (servien.remi12 at yahoo.fr) wrote: >I want to use splines to estimate a function but i want to force the
interpolation to be monotone. Is this possible with R ?
There are a few options. You can use mono.con in mgcv (see the example code in
?pcls), or smooth.monotone in fda. In both of these you have to specify the
smoothing parameter yourself. I've written (with some help from Simon Wood)
a set of scripts that fit a monontone nondecreasing regression spline with
smoothing parameter chosen by GCV taking account of the active constraints, the
approach developed by
Wood, S.N. (1994) Monotonic smoothing splines fitted by cross validation. SIAM
Journal on Scientific Computing 15:1126-1133.
You can get them at www.eeb.cornell.edu/ellner/software.html; scroll down and
look for "Gradfit".
The function you need is gcv.rssM. It isn't very efficient, so you may have
trouble if your data set is large.
Stephen P. Ellner (spe2 at cornell.edu)
Department of Ecology and Evolutionary Biology
Corson Hall, Cornell University, Ithaca NY 14853-2701