I am trying to control a binomial proportion that hovers pretty close around 0.1% - 1.0%. Since it seems like dropping from 0.1% to 0.05% is equally as severe as increasing to 0.2% (halving the occurrences, doubling the occurrences), it makes sense to me to use geometric mean / standard deviation. Also, by using geometric standard deviation, it seems I have an added benefit of not having the lower limit become less that 0.0% (an impossible value for my data).
Are there any examples that support using geometric calculations in statistical process control or is it a bad idea altogether?
