Hi Marcus
Welcome to the board
I never saw applications with geometric mean with negatives (doesn't mean it doesn't exist, of course).
Usually when you calculate the geometric mean of the returns of an investment over the years with positive and negative returns you are trying to find an equivalent average compounded return for the period.
The returns you usually use are not the percentages you add or subtract but the values you multiply to get the final amount.
Example
You invest 100
Year 1: return 20% (total: 120)
Year 2: return -5% (total: 114)
The geometric mean is calculated relative to the total (100%)
=((1+.2)*(1-.05))^(1/2)-1= 6.77%
This means that an investment with a constant return of 6.77% each period is equivalent to your investment.
You invest 100
Year 1: return 6.77% (total: 106.77)
Year 2: return 6.77% (total: 114)
Is this not your case? Please elaborate and post a sample with inputs, the logic and the expected results.
Kind regards
PGC