### Bootstrap Regression Coefficients
### Example: The hormone data. Amount in
milligrams of anti-inflammatory hormone
### remaining in 27 devices, after a
certain number of hours of wear. The devices
### were sampled from 3 different
manufacturing lots, called 1, 2, 3. Lot 3 looks
### like it had greater amounts of
remaining hormone, but it also was worn the least
### number of hours.
x <- c(1,99,25.8,
1,152,20.5,
1,293,14.3,
1,155,23.2,
1,196,20.6,
1,53,31.1,
1,184,20.9,
1,171,20.9,
1,52,30.4,
2,376,16.3,
2,385,11.6,
2,402,11.8,
2,29,32.5,
2,76,32,
2,296,18,
2,151,24.1,
2,177,26.5,
2,209,25.8,
3,119,28.8,
3,188,22,
3,115,29.7,
3,88,28.9,
3,58,32.8,
3,49,32.5,
3,150,25.4,
3,107,31.7,
3,125,28.5)
H <- matrix(x, ncol=3, byrow=T) #
create data matrix
# create the variables
lot <- H[,1]
hrs <- H[,2]
amount <- H[,3]
# plot the data ignoring lot
plot(amount,hrs)
# fit the linear model hrs = a + b*amount
+ e
hormone <- data.frame(hrs,amount)
H.lm <- lm(amount ~ hrs, hormone)
summary(H.lm)
# bootstrap the linear model
hormone.boot <- bootstrap(hormone,
coef(lm(amount ~ hrs, hormone))
summary(hormone.boot)
plot(hormone.boot)
# Next, the jackknife after the bootstrap
is used to assess the accuracy of the3 standard
# error estimates, and the influence of
each observation on these estimates.
hormone.jack <-
jack.after.bootstrap(hormone.boot,"SE")
hormone.jack
plot(hormone.jack)
|