R help - Feb 2012 - How indices calculated in package "boot"

Hi,there,

I am using R package "boot" to bootstrap. I have one question here:
does
anybody possibly know how the boot package generates the "indices"
which is
used in the statistic function?

I thought "indices = sample(data, replace=TRUE)", but when I replaced
"indices" with this command and used "boot", I got different
results.

Specifically, below are the codes for illustration.

(1) The typical way by generating indices in the package:
boot1 <- function (data, indices)
{
 d <- data[indices]
 return(d)
}
AA <- c(1:10)
require(boot)
set.seed(123)
results1 <- boot(data= AA, statistic=boot1, R=100)

(2) The alternative way by calculating "indices" myself:
boot2 <- function (data,indices)
{
 indices <- sample(data, replace=TRUE)
 d <- data[indices]
 return(d)
 }
AA <- c(1:10)
set.seed(123)
results2 <- boot(data= AA, statistic=boot2, R=100)

When I looked up using results1$t and results2$t, I had totoally different
bootstrap samples. I found this even had great impacts on the results in my
study. Does the second approach have any problem? Anyone could provide any
inputs on this? Thank you very much in advance!
Regards
Nina

	[[alternative HTML version deleted]]

In your second example the boot function will still generate its
random indicies before your internal function calls sample, so the
seed will be different when you call sample from what you originally
set it to.  If you really want to know what the boot function does,
look at its code (it does some things to try and be more efficient
which might give different results than a simpler version).

On Wed, Feb 8, 2012 at 1:45 PM, Deng Nina <nndeng at gmail.com>
wrote:
> Hi,there,
>
> I am using R package "boot" to bootstrap. I have one question
here: does
> anybody possibly know how the boot package generates the
"indices" which is
> used in the statistic function?
>
> I thought "indices = sample(data, replace=TRUE)", but when I
replaced
> "indices" with this command and used "boot", I got
different results.
>
> Specifically, below are the codes for illustration.
>
> (1) The typical way by generating indices in the package:
> boot1 <- function (data, indices)
> {
> ?d <- data[indices]
> ?return(d)
> }
> AA <- c(1:10)
> require(boot)
> set.seed(123)
> results1 <- boot(data= AA, statistic=boot1, R=100)
>
> (2) The alternative way by calculating "indices" myself:
> boot2 <- function (data,indices)
> {
> ?indices <- sample(data, replace=TRUE)
> ?d <- data[indices]
> ?return(d)
> ?}
> AA <- c(1:10)
> set.seed(123)
> results2 <- boot(data= AA, statistic=boot2, R=100)
>
> When I looked up using results1$t and results2$t, I had totoally different
> bootstrap samples. I found this even had great impacts on the results in my
> study. Does the second approach have any problem? Anyone could provide any
> inputs on this? Thank you very much in advance!
> Regards
> Nina
>
> ? ? ? ?[[alternative HTML version deleted]]
>
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.


-- 
Gregory (Greg) L. Snow Ph.D.
538280 at gmail.com