cv.grid {ePCR}

Cross-validation runs for risk predition for a grid of predetermined alpha values and their conditional lambda values

Description

Expanded Cross-Validation function to run the whole CV in the lambda/alpha grid instead of just lambda-sequence with a pre-specified alpha

Usage

cv.grid(
  alphaseq = seq(from = 0, to = 1, by = 0.1),
  seed,
  x,
  y,
  folds = 10,
  nlamb = 100,
  verb = 0,
  scorefunc,
  plot = FALSE
)

Arguments

alphaseq

Sequence of alpha values to test, which should be within [0,1] (with alpha = 0 being ridge regression, 0 < alpha < 1 being elastic net, and alpha = 1 being LASSO)

seed

Random number generation seed for reproducibility

x

Data matrix x

y

The Surv-object response y

folds

Number of folds in the cross-validation

nlamb

Number of lambda values to test in each alpha; notice that these lambda values vary conditional to alpha

verb

Level of verbosity, with 0 as silent and 1 with additional output

scorefunc

Chosen scoring function, e.g. score.cindex or score.iAUC

plot

Whether a performance should be plotted at each varying alpha-value similar to cv.alpha-plots

Value

List of matrices of cross-validation performance values over the alpha/lambda grid for mean/median/min/max/stdev of the chosen performance metric, with rows indicating various alpha-values and columns indicating lambda-values.

Examples

data(TYKSSIMU)
library(survival)
ydat <- Surv(event = yMEDISIMU[,"DEATH"], time = yMEDISIMU[,"LKADT_P"])
cvs <- cv.grid(x = xMEDISIMU, y = ydat, folds = 3, nlamb = 30, alphaseq = seq(0, 1, by=5), 
	scorefunc = score.iAUC, plot = TRUE, seed = 1)
cvs