**Date:** Mon, 15 Nov 1999 16:41:44 +0000
**Reply-To:** John Whittington <John.W@MEDISCIENCE.CO.UK>
**Sender:** "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
**From:** John Whittington <John.W@MEDISCIENCE.CO.UK>
**Subject:** Re: Statistic
**In-Reply-To:** <199911151436.OAA22628@vicar.netnames.net>
**Content-Type:** text/plain; charset="us-ascii"
At 16:20 15/11/99 +0200, dwall@QBLM.QUINTILES.COM wrote:

>Has anyone heard of a geometric CV (co-efficient of variation).
>I know what a geometric mean and SD are.
>I have some test data and it appears that the geometric CV is
>the SD of log-transformed data (without anti-logging again).
>This is then multiplied by 100 to convert to a percentage.
>Is this correct???

I can't say I've heard of a 'geometric CV' but I would be rather surprised
if it were what you have described. By analogy with a normal 'CV', one
would imagine that a geometric one would have to be expressing the
variability in terms of the mean in some fashion - whereas what you
describe is simply a function of the SD.

If you had said that it was the SD of log-transformed data DIVIDED BY the
mean of log-transformed data (both without anti-logging), then multiplied
by 100, then that would have made much more sense.

Another possibility, of course, would be to take it as SD/mean*100, using
the back-transformed (i.e. anti-logged) SD and mean - perhaps reasonable in
some senses, in as much as it is only really after back-transformation that
the SD and mean *are* really 'geometric'. Otherwise, what you'd be talking
about would be 'the CV of log-transformed data'.

I'm sure someone around will know the true/conventional answer!

Regards

John

----------------------------------------------------------------
Dr John Whittington, Voice: +44 (0) 1296 730225
Mediscience Services Fax: +44 (0) 1296 738893
Twyford Manor, Twyford, E-mail: John.W@mediscience.co.uk
Buckingham MK18 4EL, UK mediscience@compuserve.com
----------------------------------------------------------------