Date:         Thu, 3 Feb 2005 17:10:59 +0100
Reply-To:     Marta García-Granero
              <biostatistics@terra.es>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From:         Marta García-Granero
              <biostatistics@terra.es>
Organization: Asesoría Bioestadística
Subject:      Re: Confidence intervals for geometric mean
In-Reply-To:  <OF847F5196.D7D93153-ON85256F9D.004D11EC-85256F9D.004D510A@hivresearch.org>
Content-Type: text/plain; charset=ISO-8859-15

Hi Christian,

Working with GM is the same as working with the arithmetic mean of the
log(data). Compute CI for the mean of log(data), then take the antilog
of the limits.

I don't know of any nonparametric test that compares GM. If the
log(data) are roughtly normal, then run a t test on them and antilog
the limits of the CI for mean diferences (be careful with the
interpretation of the antilog results, you will get CI for the ratio
of both GM).

Take a look at these on-line articles (from British Medical Journal):

Transforming data:
http://bmj.bmjjournals.com/cgi/content/full/312/7033/770

Transformations, means and confidence intervals:
http://bmj.bmjjournals.com/cgi/content/full/312/7038/1079

The use of transformation when comparing two means:
http://bmj.bmjjournals.com/cgi/content/full/312/7039/1153

CB> I am looking some script to estimate the confidence intervals at 95% for
CB> geometric mean (GM) values. In addition, I have two groups and I am
CB> thinking to apply a non-parametric test to compare their GMs.If you have
CB> others ways to compare GM, please go ahead, welcome all ideas and comments.

HTH
 Marta                            mailto:biostatistics@terra.es