cnr: Control Net Reduction

Description

Run the Control Net Reduction Algorithm.

Usage

cnr(x, margin, n_polycoef = 20L, progress = c("cnr", "influence", "none"), ...)

Value

A cpr_cnr object. This is a list of cpr_cn objects.

Arguments

x: a cnr_cn object
margin: the margins to apply the CNR algorithm to. Passed to influence_weights.
n_polycoef: the number of polynomial coefficients to use when assessing the influence of each internal knot.
progress: controls the level of progress messaging.
...: not currently used

Details

cnr runs the control net reduction algorithm.

keep will keep the regression fit as part of the cnr\_cp object for models with up to and including keep fits. For example, if keep = 10 then the resulting cnr\_cnr object will have the regression fit stored in the first keep + 1 (zero internal knots, one internal knot, ..., keep internal knots) cnr\_cp objects in the list. The limit on the number of stored regression fits is to keep memory usage down.

Examples

Run this code


acn <- cn(log10(pdg) ~ btensor(list(day, age)
                               , df = list(10, 8)
                               , bknots = list(c(-1, 1), c(44, 53)))
         , data = spdg)
cnr0 <- cnr(acn)
cnr0
summary(cnr0)
plot(cnr0)