Construct Matrix of Resamples

DESCRIPTION:

Generate random indices for use by bootstrap.

USAGE:

samp.boot.mc(n, B) 
samp.boot.bal(n, B) 
samp.permute(n, B) 

REQUIRED ARGUMENTS:

n
sample size, samples of size n will be generated from the vector 1:n.
B
number of resamples to draw.

VALUE:

matrix in which each column is one resample.

DETAILS:

These functions are examples of samplers suitable for bootstrap. They produce a matrix of resamples from a specified vector. Each column is one set of resamples. samp.boot.mc provides simple Monte Carlo resamples. samp.boot.bal does balanced resampling in which each observation appears exactly B times in the result. samp.permute returns B columns of random permutations of 1:n.

These samplers are typically called multiple times by bootstrap, to generate indices for a block of say B=100 replications at a time; the value of B here corresponds to the block.size argument to bootstrap.

For balanced bootstrapping the individual samples are n values generated with replacement (from a vector of length nB). This produces biased results, of order O(1/B), and tends to underestimate bootstrap standard errors and produce confidence intervals which are too narrow.

SEE ALSO:

.

EXAMPLES:

samp.boot.mc(4, 2)