bootstrap and other
resample objects.
resample objects, or create it if necessary.
resampGetIndices extracts a matrix of resampling indices from a resampling
object, or creates it from scratch if necessary.
resampGetArgument extracts arguments used to create resample
objects.
resampGetIndices(object, frame.eval = <<see below>>)
resampGetArgument(object, argument, checkObject = <<see below>>,
frame.eval = <<see below>>)
resample object (not all such objects are appropriate for
these functions).
object.
TRUE, look first in
object for a component with name equal to
argument. This is the default action unless
argument equals
"B", in
which case the value is recreated from the original call to the
resample function.
object can be found.
You need to specify this if objects can't be found by their
original names, or have changed; see
.
resampGetIndices returns a matrix of indices.
resampGetArgument returns the value of
argument,
evaluated in an appropriate context.
The context is important in two cases. First, if the object were created
inside a function and made use of objects defined in that function,
then those objects, or identical copies, should be accessible;
this is done by letting
frame.eval be the number of the frame
of that function, or a new frame containing copies of those objects.
Second, if the
data argument to the resample function is a data
frame, that data frame, then arguments like
group and
subject are
searched for there first.
resampGetIndices just extracts an
indices component if one exists
in the object. Otherwise it creates a new call to the original function, with
the
statistic set to a function that returns
NULL.
Results will be incorrect if the sampler is adaptive, where indices
depend on the value of the statistic returned in earlier replications.
If
checkObject=TRUE,
resampGetArgument first checks
object for a
component with name equal to
argument and, if found, returns that
value. If the
component is not found (or if
checkObject=FALSE), the argument is
reconstructed from the original call. Symbolic references in the call
(as in the case where the bootstrap
data argument is a data frame,
and
group is the name of a column of
data) are evaluated so
that the actual value of the argument is returned. Note that this may
cause unintended results in the case
argument="statistic": if
statistic is an expression, the expression is evaluated (if
possible) and that value returned. If you want the unevaluated
expression for an argument, access the component directly, using, for
example,
object$call$statistic.
The special default action when
argument="B" addresses the case when
bootstrap argument
B is a vector. In this case, the value stored
in the bootstrap object is not the vector, but
sum(B),
the total number of resamples. The original vector
B can be
recreated from the call.
# get indices boot <- bootstrap(stack.loss, mean, B=50) resampGetIndices(boot) # get group argument df <- data.frame(x=stack.loss, a=rep(1:3, 6:8)) boot <- bootstrap(df, mean(x), group=a, B=50) resampGetArgument(boot, "group")