I have an ensemble of 3 reanalysis. What is the correct method to calculate the spread of the ensemble? I doubt between to apply inter-reanalysis standard deviation or the range (max - min) of the ensamble. Thanks in advance
this is a hard problem given that you have only an "ensemble" of three members. Did I understand it well?
The statistics will suffer from the small ensemble size. The standard deviation (ensemble spread) can be normalized by dividing by N-1=2 instead of N=3 to obtaine less biased results. Not sure whether it helps. If you want some reliable results you have to compute these statistics over a quite large number of cases.
Thanks so much by your answer. I know that is a difficult question this is why I asked for help to the ResearchGate community. I know that a small ensemble size always have troubles in a statitstical analysis. So as you recommend, I am going to use the standard deviation with the Bessel's correction (N-1, instead N) to calculate the ensemble spread.