Jackknife 본문
#jackknife - 1
theta.hat <- cor(dat$newpatch, dat$oldpatch)
n <- nrow(dat);n
theta.hat.minus <- numeric(n)
for(i in 1:n){
theta.hat.minus[i] <- cor(dat$newpatch[-i], dat$oldpatch[-i])
}
theta.hat.minus
bias.jack <- (n-1)*(mean(theta.hat.minus) - theta.hat);bias.jack
theta.jack <- theta.hat - bias.jack;theta.jack
theta.hat.star.minus <- n*theta.hat - (n-1)*theta.hat.minus
mean(theta.hat.star.minus)
se <- sqrt((n-1)*mean((theta.hat.minus - mean(theta.hat.minus))^2));se
#jackknofe - 2
library(bootstrap)
dat <- law;dat
theta.hat <- cor(dat$LSAT, dat$GPA)
n <- nrow(dat);n
theta.hat.minus <- numeric(n)
for(i in 1:n){
theta.hat.minus[i] <- cor(dat$LSAT[-i], dat$GPA[-i])
}
bias.jack <- (n-1)*(mean(theta.hat.minus) - theta.hat)
se.jack <- sqrt((n-1)*mean((theta.hat.minus - mean(theta.hat.minus))^2))
'Applied > Statistical Computing' 카테고리의 다른 글
Permutation Test (1) | 2022.12.01 |
---|---|
Chapter 14.5 ) The EM - Algorithm (0) | 2022.11.22 |
Chapter 14 - 1 ) Quasi-Newton Algorithm (0) | 2022.11.18 |
Chapter 14 - 0 ) Optimization Introduction (0) | 2022.11.17 |
Bootstrap estimate of bias (0) | 2022.11.13 |
Comments