bootstrap.
samp.boot.mc(n, B) samp.boot.bal(n, B) samp.permute(n, B)
n will be generated from the vector 1:n.
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.
samp.boot.mc(4, 2)