This query comes up regularly on allstat. I have put an answer to in,
taken from my book An Introduction to Medical Statistics, on my website,
address below.
Martin
Jay Bryant wrote:
>
> hi.
>
> i have n samples x1, x2,...,xn of a random variable x
> of an unknown distribution. i want to estimate the 95%
> ci of the 100*alpha percentile of x (for example the
> ci for the 5% percentile of x). i thought of using the
> bootstrap, but is this mathematically justifiable
> since the statitic "100*alpha percentile" is not
> really a functional statistic?
> any suggestions and ideas would be greatly
> appreciated. thanks.
>
> regards,
>
> j.
>
> __________________________________________________
> Do You Yahoo!?
> LAUNCH - Your Yahoo! Music Experience
> http://launch.yahoo.com
--
**************************************************
Martin Bland
Prof. of Medical Statistics
St. George's Hospital Medical School
London SW17 0RE
[log in to unmask]
http://www.mbland.ac.uk/
044 (0)20 8725 5492
**************************************************
|