| gam2objective {mgcv} | R Documentation |
Estimation of GAM smoothing parameters is most stable if
optimization of the UBRE or GCV score is outer to the penalized iteratively
re-weighted least squares scheme used to estimate the model given smoothing
parameters. These functions evaluate the GCV/UBRE score of a GAM model, given
smoothing parameters, in a manner sutable for use by optim or nlm.
Not normally called directly, but rather service routines for gam.outer.
gam2objective(lsp,args,printWarn=FALSE) gam2derivative(lsp,args) gam3objective(lsp,args)
lsp |
The log smoothing parameters. |
args |
List of arguments required to call gam.fit2. |
printWarn |
Should gam.fit2 print some warnings? Used to
suppress warnings that are only of interest for the final fitted model, until
that model is reached. |
gam2objective and gam2derivative are functions suitable
for calling by optim, to evaluate the GCV/UBRE score and it's
derivatives w.r.t. log smoothing parameters.
gam3objective is an equaivalent to gam2objective, suitable for
optimization by nlm - derivatives of the GCV/UBRE function are
calculated and returned as attributes.
The basic idea of optimizing smoothing parameters `outer' to the P-IRLS loop was first proposed in O'Sullivan et al. (1986).
Simon N. Wood simon.wood@r-project.org
O 'Sullivan, Yandall & Raynor (1986) Automatic smoothing of regression functions in generalized linear models. J. Amer. Statist. Assoc. 81:96-103.
http://www.stats.gla.ac.uk/~simon/