controlVariates(x, subset.covariates = 1:p, moments = 2, ..., L, group, treatment, frame.eval)
bootstrap
object.
x
or computed if needed;
you need to supply it only if it cannot be computed.
Similarly for
group
,
treatment
, and
frame.eval
.
"controlVariates.bootstrap"
,
similar to
x
but with:
Mean
,
Bias
, and
SE
, calculated using weights
selected to match one or two theoretical moments of a linear approximation
to the statistic (or columns of a multivariate statistic) being
bootstrapped.
B
, the number of bootstrap samples.
See the detailed explanation in .
If the original bootstrap object
x
contained
weights
then this
is the product of those weights with adjustments to match the specified
moments
If
x
is a
bootstrap2
object, weights and estimates are only
computed for the main object, not the two contained bootstrap objects.
Hesterberg, T.C. and Nelson, B.L. (1998), "Control Variates for Probability and Quantile Estimation," Management Science, 44(9), 1295-1312.
Hesterberg, T.C. (1996), "Control Variates and Importance Sampling for Efficient Bootstrap Simulations," Statistics and Computing 6(2), 147-157.
boot <- bootstrap(1:9, mean, seed=0, save.indices=T) boot2 <- controlVariates(boot) boot boot2 # note that Mean = 5, Bias = 0, and # SE = sqrt( mean( (1:9-5)^2 ) / 9 )