On 10 Sep 2001 15:37:18 -0700, [EMAIL PROTECTED] (JD Kronicz)
wrote:
> Hi-
> I am a human health risk assessor. I am trying to calculate the 95%
> UCL of a lognormal distribution. The data sets I have all have about
> 350 samples. I can't find a lookup table for the H-Statistic for more
> than 100 samples. If anyone can point me in the right direction I
> would greatly appreciate it.
> Thanks-JDK
Like Alan, I never heard of this before, either.
I searched with Google on "H-statistic" and most of the hits were
for Durbin's test on autocorrelation, which is not the right one.
Everyone relevant seems to refer to RA Gilbert's book, 1987.
Is that where there is a table that only goes to 100?
Here is a reference that raises objections to the H-statistic method
(since real distributions are, too often, not log-normal enough)
and describes a bootstrapping alternative:
http://www.state.ak.us/dec/dspar/csites/guidance/tech_memos/bootstrap.pdf
Hope this helps.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
=================================================================
Instructions for joining and leaving this list and remarks about
the problem of INAPPROPRIATE MESSAGES are available at
http://jse.stat.ncsu.edu/
=================================================================