gencve (version 0.3)

regal: Regression EPE for All Implemented Methods

Description

Determine EPE for many regression methods.

Usage

regal(X, y, MaxIter = 1000, d = "default", NCores = 1, plotBest = 6, 
	verboseQ = FALSE)

Arguments

X

input matrix of dimension n-by-p with p

y

output vector

MaxIter

Number of CV iterations.

d

Size of hold-out sample.

NCores

Number of cores to use for parallel processing.

plotBest

Number of EPE's to include on plot

verboseQ

True, display progress, otherwise silent. When running R in Windows, the usual default is output buffering which means you will not see the extra output generated from verboseQ=TRUE until after regal() has finished. To see the output while this function is running you need to turn output buffering off. This can be done with the short-cut Ctrl-W. Another way to do this is to use the R Gui. Select Misc and the click on buffered output.

Value

A barplot is produced and matrix returned with rows corresponding to method and columns containing EPE, sd(EPE), snr and two correlation estimates between forecast and true value.

See Also

gcv

Examples

#about 200 seconds
## Not run: 
#   data(prostate)
#   X <- as.matrix.data.frame(prostate[,-9])
#   y <- prostate[,9]
#   system.time(m<-regal(X, y, MaxIter=1000, d=10, NCores=8, verboseQ=TRUE))[3]
#   ind <- rev(order(m[,6]))
#   dotchart(m[ind,6], pch=19, cex=0.5, bg=rgb(1,1,0,0.4),
#        color="blue", main="CPU times")
# ## End(Not run)