an extension to the geometric standard deviation query


I am interested in calculating reliable change indices that take the
form of (x2-x1)/sdchange.

for several of my variables there is violation of normality that can be
corrected with log10 transformation.

Am i correct that the sd of this value needs to be antilogged to be
meaningful (creating a geometric SD), and in this case, is it valid to
use this antilogged value in the rci calculation?

I'd be grateful for any feedback

.