Non-Parametric Bootstrap

Refer to Bootstrap

We perform the non-parametric bootstrap

x.nonparbs1 <- sample(x, size=10, replace=TRUE)
 
mean(x.nonparbs1, trim=.2)
 
x.nonparbs10000 <- replicate(10000,
 
mean(sample(x, size=10, replace=TRUE), trim=.2))
 
sd(x.nonparbs10000)

In this case, the outliers can appear several times in the resample meaning that the standard error could be inflated.

Recall $\bar{X}=\frac{\sigma }{\sqrt{n}}$