Following up on the 'Nuke Geometric Returns' post I tried sketch out possible equations for the variance of geometric returns.
I got good looking numerical answers, but the formula looked very odd.
Then spent a few hours searching for the answer with Google.
Thankfully Gordon on Quant Stack Exchange helped out!
Nice approach.
Now, I have to avoid the temptation to figure out covariance and other analytics - have too much on my plate already.
In any case, the original critique still stands.
Never use geometric returns for any time series analysis! Invariably your analysis will be incorrect.