[R] [OT?] Question on bootstrapping

Matthias Gondan matthias-gondan at gmx.de
Sun Feb 3 11:45:32 CET 2008


(little correction below)

> Dear list,
>
> This is off-topic, but perhaps there is an expert out there who can help 
> me in
> a bootstrapping test.
>
> I have two samples A, B from, say, a normal distribution. A third sample R
> is the vector of pairwise minima of  the same random variables:
>
> A = rnorm(100)
> B = rnorm(100)
>
> A2 = rnorm(100)
> B2 = rnorm(100)
> R = pmin(A2, B2)
>
> For some purposes, I want to do some bootstrapping. From A and B, this
> should not be a problem:
>
> sa = sample(A, replace=T)
> sb = sample(B, replace=T)
>
> But for the minima distribution, I have the "feeling" (sorry) that the 
> bootstrapped
> values are systematically too high:
>   

too high = compared to the random variable R, as defined above

> a2 = sample(A, replace=T)
> b2 = sample(B, replace=T)
> sr = pmin(a2, b2)
>
> Can anybody give me a hint whether my feeling is correct, and perhaps 
> indicate
> some literature in which such issues are handled?
>
> Best wishes,
>
> Matthias
>
>



More information about the R-help mailing list