Strictly speaking, it is incorrect to use standard deviation (or average as well) to describe the data as you are operating with frequencies and not the primary numbers. It might be less confusing to use median and some simple non-parametric comparison because frequencies do not follow normal distribution.
I took a liberty and checked Calv1n’s data as an example to see whether they match normal distribution and they don’t according to Shapiro-Wilk test (P=3.18E-7) or Anderson-Darling test (P=3.19E-13). I’m using real-statistics addon for Excel.
For comparison, it might be easier to simulate data with one or another assumed primary value and check which one of these simulations is a better fit for actual data.
In general, this type of frequency data for cumulative probabilities do not follow normal distribution, working as intended in this case as well.
