Additional arguments provided by ssr are subset, scale, limnla and control. subset selects a subset of the data for fitting. scale, if T (true), scales all covariates in the rk argument into . It is recommended that scaling be done before fitting. limnla, a vector of length 1 or 2, sets the searching limits for on log10 scale when fitting a univariate smoothing spline model (). One may fix the smoothing parameter by setting the length of limnla to 1. For example, one may fit a cubic spline with by
ssr(y~t, rk=cubic(t), limnla=log10(0.01))
The control option specifies several control parameters used in RKPACK and GRKPACK. See ssr.control for details.
Generic utility functions supporting ssr include summary, plot, deviance, residuals, and hat.ssr, in addition to the anova and predict function discussed in previous sections. The summary function provides the basic description of a ssr fit. plot produces diagnostic plots for a ssr object. hat.ssr returns the hat matrix of a spline fit. Note that the full name hat.ssr should be used since the name ``hat'' has been utilized for another purpose. See the help files in the package for more detailed descriptions.