crossvalidate.pim: Crossvalidate a glmnet PIM

In pimold: Probabilistic Index Models

Description

Crossvalidate a glmnet PIM

Usage

crossvalidate.pim(pimob, method = c("fullsplit", "semisplit", "naive"),
  type.measure = c("mse", "deviance", "class", "auc", "mae"), nfolds = 10,
  foldid, weights, ..., include.extrainfo = FALSE, verbosity = 0)

Arguments

pimob	Object of class pim. Note: it should really be of class glmnetpim in the current implementation.
method	"fullsplit" splits pseudo-observations over folds, "semisplit" takes the nonvalidating pseudo-observations together for the fit for each fold, "naive" simply crossvalidates over the pseudo-observations (note: typically, this is not correct!)
type.measure, nfolds, foldid, weights	See cv.glmnet
include.extrainfo	If TRUE (not the default), the return value contains two extra items, foldid and foldFits, that may be used to reassess the crossvalidation value
verbosity	The higher this value, the more levels of progress and debug information is displayed (note: in R for Windows, turn off buffered output)
...	Passed on to cv.glmnet (or .cvpo.glmnet).

Value

An object of class cv.pim, of class cv.glmnet, and depending on the type of linkfunction (see cv.glmnet) some more classes. It holds all the items of the original pimob and all necessary items for a cv.glmnet object (so the relevant S3 methods like print and plot will work for them.)

See Also

pim cv.glmnet

Examples

set.seed(1)
pen.N<-100
pen.noisep<-50
pen.noisemat<-matrix(rnorm(pen.N*pen.noisep), nrow=pen.noisep)
pendta<-data.frame(y=rnorm(pen.N), x=factor(sample(2, pen.N, replace=TRUE)), pen.noisemat)
pendta$y[pendta$x=="2"]<-pendta$y[pendta$x=="2"]+1
colnames(pendta)[(1:pen.noisep)+2]<-paste("X", 1:pen.noisep, "X", sep="")
pen.formula<-paste("y~", paste(c("F(x)", setdiff(colnames(pendta), c("x", "y"))), collapse="+"), sep="")

penpim<-pim(as.formula(pen.formula), data=pendta, link="identity", poset=noselfposet, 
	estimator=estimator.glmnet(), varianceestimator=NULL, keep.data=TRUE, verbosity=0,
	interpretation="regular")
cv.penpim.naive<-crossvalidate.pim(penpim, method="naive")
cv.penpim<-crossvalidate.pim(penpim, method="fullsplit")
cv.penpim.ps<-crossvalidate.pim(penpim, method="semisplit")

pimold documentation built on May 2, 2019, 5:50 p.m.