I can't speak to financial convention, since I don't work in that area.

However, for most things that you would want to do with a standard
deviation, back-transforming the standard deviation itself would not be
useful. For instance, if you had to transform your data to achieve
approximate normality, you would calculate confidence intervals, etc on the
thransformed scale and back-transform those limits rather than the standard
deviation itself.

Jerry

"joeu2004" wrote:

On Sep 13, 10:56 am, Jerry W. Lewis
wrote:
If you mean as defined athttp://en.wikipedia.org/wiki/Volatility_%28finance%29
you could use the array formula (committed with Ctrl-Shift-Enter)
=STDEV(LOG(data))*SQRT(tradingDaysPerYear)
to get annualized volatility.

I notice that you (and the Wiki entry) do not convert that back to a
std dev of the returns (i.e. exp(stdev(ln(data))...)). Is that
conventional? If so, would you also leave the mean in terms of the
log return (e.g. average(log(data)))?

I also wonder what financial software really wants when it asks us to
input, variously, "volatility" or "std dev" along with the "mean".