addSamples.bootstrap(object, B.add = 100, sampler.prob = NULL, ..., frame.eval = <<see below>>) addSamples.parametricBootstrap(object, B.add = 100, ..., frame.eval = <<see below>>) addSamples.smoothedBootstrap(object, B.add = 100, ..., frame.eval = <<see below>>)
B
argument in
.
length(B.add)
; probabilities to be used for
importance sampling. See
for more information.
object
can be found.
You need to specify this if objects can't be found by their
original names, or have changed; see
.
object
, with additional replicates.
The component
B
of the result is the sum of the total previous
samples (component
B
of the old object)
and the sum of elements of
B.add
.
The component
call$B
of the result is a vector concatenating the
previous
call$B
and
B.add
.
If the original call (to
,
,
or
) was interrupted,
incomplete results (e.g. saved in
.bootstrap.partial.results
)
may be extended with
addSamples
, but
you must manually change the
seed.end
component
to a legal
seed
argument first.
If
object
is a bootstrap object,
components
L
and
Lstar
of the result are the same as for
object
:
they are not updated to reflect the new samples.
For an annotated list of functions in the package, including other high-level resampling functions, see: .
bfit <- bootstrap(stack.loss, median, B = 1000) bfit2 <- addSamples(bfit, B.add = 400) bfit2 # The call has "B = c(1000, 400)" # Demonstrate adding samples after an interrupt # interrupt this next call after 200 replications or so bfit3 <- bootstrap(stack.loss, median, seed = bfit$seed.start) bfit3 <- .bootstrap.partial.results set.seed(bfit$seed.start) dummy <- samp.bootstrap(bfit3$n, bfit3$B) # to update .Random.seed bfit3$seed.end <- .Random.seed # manually change seed.end bfit4 <- addSamples(bfit3, B.add = 1400-bfit3$B) all.equal(bfit2, bfit4) # same except for call, actual.calls all.equal.excluding(bfit2, bfit4, excluding= c("call", "actual.calls")) # T # parametricBootstrap bfit <- parametricBootstrap(iris[,1,1], mean, rsampler = rnorm, args.rsampler = list(mean = mean(iris[,1,1]), sd = sqrt(var(iris[,1,1]))), B = 600) bfit2 <- addSamples(bfit, B.add = c(300,100)) bfit3 <- update(bfit, B=c(600, 300, 100), seed=bfit$seed.start) all.equal(bfit2, bfit3) # same except for call, actual.calls all.equal.excluding(bfit2, bfit3, excluding= c("call", "actual.calls")) # T # smoothedBootstrap bfit <- smoothedBootstrap(stack.loss, mean, B = 600) bfit2 <- addSamples(bfit, B.add = c(300,100)) bfit3 <- smoothedBootstrap(stack.loss, mean, seed = bfit$seed.start, B = 1000) all.equal(bfit2, bfit3) # same except for call, actual.calls all.equal.excluding(bfit2, bfit3, excluding= c("call", "actual.calls")) # T