### Bootstrap t CI for the trimmed mean of the time on target data

ttarg <- c(77,87,88,114,151,210,219,246,253,262,296,
	299,306,376,428,515,666,1310,2611)
	
gamma <- 0.20		# proportion to be trimmed on both sides

tmuhat <- mean(ttarg,trim=gamma)
tmuhat

ttmean2 <- function(X){
	x <- sort(X) 	# sort X for Winsorizing
	n <- length(x)
	a <- mean(x,trim=gamma)
	den <- n*(1 - 2*gamma)^2
	tt <- floor(n*gamma)
	lower <- x[tt-1]	# l for low, rep l in wv
	upper <- x[n-tt] 	# u for up, rep u in wv
	wv <- c(rep(lower,time=tt),x[(tt+1):(n-tt)], rep(upper,time=tt))
	b <- (a-tmuhat)/sqrt(var(wv)/den)
	c(a,b)	# returns the two values
}

ttmean2.boot <- bootstrap(ttarg,ttmean2,B=2500)

print(ttmean2.boot)
plot(ttmean2.boot)

ttmean2.boot$estimate[1,3]		# the SE of the trimmed means

setmn <- ttmean2.boot$estimate[1,3]	# location of the SE in the bootstrap data.frame
setmn

ttmean2.boot.emp <- limits.emp(ttmean2.boot)
ttmean2.boot.emp

finv05 <- ttmean2.boot.emp[2,2]
finv05

finv95 <- ttmean2.boot.emp[2,3]
finv95

ttmean2.btci <- c(tmuhat - (setmn*finv95), tmuhat - (setmn*finv05))
ttmean2.btci