Is this a suitable way to summarize parameter distributions?


New Member
Hello everyone,

I have the following situation: I have a set of multiple environmental models with multiple parameters. I have now run these with multiple time steps to observe the change in parameter probability distribution, as you change the time step. I have done this with a Monte Carlo approach. In other words, I have thousands of parameter values and hence need a very practical statistic to summarize these.

As for any one parameter, I am interested in the CHANGE of the value (when using different time steps), I basically need to a statistic that evaluates this change. So one option could be something like using as Kolmogorov-Smirnov test, but could this work alternatively:

Would it be valid to compare the coefficient of variation at a certain reference time step (i.e. st dev of ref/mean of ref) with the change in mean at the new time steps (i.e. delta mean/mean of ref)? Hence, if this change is large than the CoV, then there is a significant variation in the distribution, otherwise not... Does this sound plausible?

Help is highly appreciated. Thank you.