randomSample(x, size = numRows(x), prob = NULL, full.partition = "none")
n (the number of observations in
x).
n, giving selection probabilities.
If the elements of prob do not sum to one, they are normalized.
The default
NULL indicates equal probabilities.
"first",
"last", or
"none"; Return the initial
(if
"first") or final (if
"last")
size elements of a full
sample of size
n. If
"none", do not generate a full sample.
Valid only if
size < n; ignored otherwise. See
for details.
This uses
balancedSample to sample indices, then subscripts the
corresponding observations from
x. The result is similar conceptually
to
sample(x, size, prob), except that this samples rows of matrices,
uses a faster sampling algorithm,
and samples with unequal probabilities correctly.
If
size>n, or
size*max(prob)>1, then sampling is done with "minimal
replacement"; see
balancedSample.
randomSample(stack.loss, size=10) # 10 elements, without replacement randomSample(stack.loss) # permutation of the original 21 elements randomSample(stack.loss, size=42) # each observation twice, random order randomSample(stack.loss, size=10, prob=c(1:21))