Summary Method for Bootstrap Objects

DESCRIPTION:

Produces a summary for bootstrap objects. This calculates percentiles and BCa confidence limits for the bootstrap distributions of parameters, and correlations of parameter replicates if applicable. The print method displays the same results as print.resamp, plus these quantities.

USAGE:

summary.bootstrap(object, probs=c(0.025, 0.05, 0.95, 0.975), 
                  frame.eval = <<see below>>, narrow = F, ...) 

REQUIRED ARGUMENTS:

object
object of class .

OPTIONAL ARGUMENTS:

probs
probability levels at which to calculate percentiles and confidence limits.
frame.eval
frame where the data and other objects used when creating object can be found. You need to specify this if objects can't be found by their original names, or have changed; see .
narrow
logical, determines how quantiles are computed. If narrow=FALSE (the default), then intervals are wider and coverage probabilities are usually more accurate than with narrow=TRUE.
...
optional arguments to pass to (for example, z0, acceleration, and group.size), or (for example, t.adjustment, initial).

VALUE:

a list containing selected components of object ( call, B, observed, estimate), plus percentiles limits.emp, BCa confidence intervals , and (if observed has more than one element) the correlation matrix correlation for the bootstrap replicates.

Tilting limits, , are also computed, if the bootstrap statistic is "functional" (the observed value is the same as the value computed with equal weights) and if either the bootstrap object contains component L (empirical influence values) or if the size of the original bootstrap sample is less than 100.

DETAILS:

The print method prints the results in a nicely formatted manner.
To get percentiles of the bootstrap distribution without computing BCa confidence intervals, call rather than summary.

NOTE:

If acceleration is estimated using jackknife, the statistic must allow samples of reduced size; see .

SEE ALSO:

, , , , , , .

EXAMPLES:

# statistic is not functional, tilting limits not computed 
summary(bootstrap(stack.loss, var))   
# statistic is functional, tilting limits are computed 
summary(bootstrap(stack.loss, var(stack.loss, unbiased = F)))