conffuns.from.bcaboot: Construct Confidence Functions from an Output of bcaboot

In ddarmon/clp: R Package for Computing Confidence Functions to Accompany CLP

Description

Construct confidence functions from the output of bcaboot. This is a helper function to make constructing ones own bootstrap confidence functions straightforward.

Usage

conffuns.from.bcaboot(bc)

Arguments

bc	an output from bcaboot

Value

A list or list of lists containing the confidence functions pconf, dconf, cconf, and qconf for the parameter associated with the statistic(s) used with bcaboot.

References

Tore Schweder and Nils Lid Hjort. Confidence, likelihood, probability. Vol. 41. Cambridge University Press, 2016.

Bradley Efron and Trevor Hastie. Computer Age Statistical Inference. Vol. 5. Cambridge University Press, 2016.

Examples

# Reproduce BCa confidence density from Figure 11.7
# of *Computer Age Statistical Inference*.

scor <-
read.table('https://web.stanford.edu/~hastie/CASI_files/DATA/student_score.txt',
header = TRUE)

statistic <- function(data, id = 1:nrow(data), ...){
  dat <- data[id, ]

  Sigma <- cov(dat)*((nrow(dat)-1)/nrow(dat))

  lams <- eigen(Sigma, symmetric = TRUE, only.values = TRUE)$values

  return(lams[1])
}

ran.gen <- function(data, mle){
  MASS::mvrnorm(n = nrow(data), mle$mu, mle$Sigma)
}

bc <- bcaboot(data = scor, statistic = statistic, B = 8000, sim = "parametric", ran.gen = ran.gen,
              mle = list(mu = colMeans(scor),
                         Sigma = cov(scor)*((nrow(scor)-1)/nrow(scor))))

lam.conf <- conffuns.from.bcaboot(bc)

display.cconf(lam.conf, xlab = 'Largest Eigenvalue')
display.dconf(lam.conf, xlab = 'Largest Eigenvalue')

lam.conf$qconf(c(0.025, 0.975))

ddarmon/clp documentation built on Jan. 25, 2021, 6:22 p.m.