mgcv.control {mgcv} | R Documentation |
This is an internal function of package mgcv
which allows control of the numerical
options for fitting a generalized ridge regression problem using routine mgcv
.
mgcv.control(conv.tol=1e-7,max.half=20,target.edf=NULL,min.edf=-1)
conv.tol |
The convergence tolerance. |
max.half |
successive step halvings are employed if the Newton method and then the steepest descent backup fail to improve the UBRE/GCV score. This is how many to use before giving up. |
target.edf |
If this is non-null it indicates that cautious optimization should be used, which opts for the local minimum closest to the target model edf if there are multiple local minima in the GCV/UBRE score. |
min.edf |
Lower bound on the model edf. Useful for avoiding numerical problems at high smoothing parameter values. Negative for none. |
Simon N. Wood simon.wood@r-project.org
Gu and Wahba (1991) Minimizing GCV/GML scores with multiple smoothing parameters via the Newton method. SIAM J. Sci. Statist. Comput. 12:383-398
Wood, S.N. (2000) Modelling and Smoothing Parameter Estimation with Multiple Quadratic Penalties. J.R.Statist.Soc.B 62(2):413-428
http://www.stats.gla.ac.uk/~simon/