Logo Questions Linux Laravel Mysql Ubuntu Git Menu
 

Values of bootstrap statistics

I want to get the values of bootstrap statistics (original, bias and error) into a separate list - but I cannot figure out how to do that.

Here's an example:

> library(boot)
> set.seed(123)
> mean.fun <- function(data, idx) { mean(data[idx]) }
> data <- boot(data=rnorm(100), statistic=mean.fun, R=999)
> names(data)
 [1] "t0"        "t"         "R"         "data"     
 [5] "seed"      "statistic" "sim"       "call"     
 [9] "stype"     "strata"    "weights"  
> data
ORDINARY NONPARAMETRIC BOOTSTRAP


Call:
boot(data = rnorm(100), statistic = mean.fun, R = 999)


Bootstrap Statistics :
    original   bias    std. error
t1* 0.09040591 0.004751773  0.08823615

Now, instead of text I want the actual values. Apparently data$t0 is the "original" but I don't see how to get the values for bias and error.


Also, since typing a function name gives you its code, I typed boost in R and copied a snippet from the source code, and tried to search it on my local R installation. But couldn't find anything. Why, shouldn't R grab that source code from a local storage?

like image 787
andreister Avatar asked Nov 12 '22 16:11

andreister


1 Answers

The std.error and bias are not stored as a part of the boot object. It is calculated on the fly (see: https://stat.ethz.ch/pipermail/r-help/2011-July/284660.html)

In your case, try:

mean(data$t) - data$t0
sd(data$t)
like image 177
jokel Avatar answered Nov 15 '22 06:11

jokel