I would like to bootstrap my data to get a confidence intervals around a mean for a vector different lengths. For example with the code below I calculate y based on my vector x and then apply bootstrap to get CI.
set.seed(001)
x <- rnorm(length(iris$Sepal.Length),iris$Sepal.Length,0.05*iris$Sepal.Length)
mydata<-data.frame(x=x,y=pi*x^2)
library(boot)
myboot<-boot(mydata$y, function(u,i) mean(u[i]), R = 999)
boot.ci(myboot, type = c("perc"))
My question is how can I compute the bootstrapped mean CI for different size of x like for 3-4, 4-5, 5-6, 6-7, 7-8?
How about this:
set.seed(001)
x <- rnorm(length(iris$Sepal.Length),iris$Sepal.Length,0.05*iris$Sepal.Length)
mydata<-data.frame(x=x,y=pi*x^2)
mydata$x_cut <- cut(x, breaks=c(3,5,6,7,9))
boot_fun <- function(data, inds){
tmp <- data[inds, ]
tapply(tmp$y, tmp$x_cut, mean)
}
library(boot)
myboot<-boot(mydata, boot_fun, R = 999, strata=mydata$x_cut)
boot.ci(myboot, type = c("perc"))
cis <- sapply(1:4, function(i)boot.ci(myboot, type="perc", index=i)$percent)
colnames(cis) <- names(myboot$t0)
cis <- cis[4:5, ]
rownames(cis) <- c("Lower", "Upper")
cis <- t(cis)
cis
# Lower Upper
# (3,5] 67.79231 72.81593
# (5,6] 92.25999 97.25919
# (6,7] 124.65315 130.88061
# (7,9] 167.58324 183.88702
for BCa intervals, use:
boot.ci(myboot, type = c("bca"))
cis <- sapply(1:4, function(i)boot.ci(myboot, type="bca", index=i)$bca)
colnames(cis) <- names(myboot$t0)
cis <- cis[4:5, ]
rownames(cis) <- c("Lower", "Upper")
cis <- t(cis)