convert a geometric Sd to arithmatic given only the geo SD statistic.

[palmerr]

Hi all. Can you help or point me in a direction with a seemingly simple conversion?

I need to convert a geometric standard deviation into the arithmetical SD.

In a published manuscript the aritmatic mean, the geometric mean and geometric SD is reported. I am trying to compute a coeffienct of variation (to do a power analysis in SAS for a 2 sample t-test from a log distribution). The program asks for a coefficient of varaition (SD/arith mean)*100, when comparing mean ratios.

but all I have is the geoSD and the paper doesnt give the aritmatic SD, is there a straightforward conversion from geo SD to SD given.

thanks, ray

[P J L]

Hi,

Sorry it's probably too late for you now, but I just wanted to point out that you don't need to convert to arithmetic SD - you can get straight from geometric SD to CV for lognormal data with the following:

SD on log scale = ln(geoSD) = s
CV = 100*sqrt(exp(s^2)-1)

See p1960 of the following ref: Statist. Med. 2004; 23:1921–1986 (DOI: 10.1002/sim.1783)

Regards,
Pete