trainind: Generate Index of Training Samples

Description

Generate index of training samples. The sampling scheme includes leave-one-out cross-validation (loocv), cross-validation (cv), randomised validation (random) and bootstrap (boot).

Usage

trainind(cl, pars = valipars())

Value

Returns a list of training index.

Arguments

cl: A factor or vector of class.
pars: A list of sampling parameters for generating training index. It has the same structure as the output of valipars. See valipars for details.

Author

Wanchang Lin

Examples

Run this code

## A trivia example
x <- as.factor(sample(c("a","b"), 20, replace=TRUE))
table(x)
pars <- valipars(sampling="rand", niter=2, nreps=4, strat=TRUE,div=2/3)
(temp <- trainind(x,pars=pars))
(tmp  <- temp[[1]])
x[tmp[[1]]];table(x[tmp[[1]]])     ## train idx
x[tmp[[2]]];table(x[tmp[[2]]])
x[tmp[[3]]];table(x[tmp[[3]]])
x[tmp[[4]]];table(x[tmp[[4]]])

x[-tmp[[1]]];table(x[-tmp[[1]]])   ## test idx
x[-tmp[[2]]];table(x[-tmp[[2]]])
x[-tmp[[3]]];table(x[-tmp[[3]]])
x[-tmp[[4]]];table(x[-tmp[[4]]])

# iris data set
data(iris)
dat <- subset(iris, select = -Species)
cl  <- iris$Species

## generate 5-fold cross-validation samples
cv.idx <- trainind(cl, pars = valipars(sampling="cv", niter=2, nreps=5))

## generate leave-one-out cross-validation samples
loocv.idx <- trainind(cl, pars = valipars(sampling = "loocv"))

## generate bootstrap samples with 25 replications
boot.idx <- trainind(cl, pars = valipars(sampling = "boot", niter=2,
                                           nreps=25))

## generate randomised samples with 1/4 division and 10 replications. 
rand.idx <- trainind(cl, pars = valipars(sampling = "rand", niter=2, 
                                           nreps=10, div = 1/4))