I want to use the eFast method in GlobalSensitivity.jl but where parameters are input as distributions rather than ranges (param_range asks for a vector of tuples). For example, I’d like to specify that a parameter be sampled from a normal distribution with a given mean and variance rather than from between an upper and lower range with equal probability (which would be the same as a uniform distribution). Is this possible with this package or another?

Can you open an issue? Global sensitivity analysis can be defined with inputs as distributions, in general there is a way to say “standard” GSA is an analysis with uniform distribution inputs, but that can be generalized.

It turns out there is already an open issue for this from February: Allow to give Distributions for `param_range` · Issue #99 · SciML/GlobalSensitivity.jl · GitHub. I didn’t want to make a duplicate issue so I just upvoted and added a comment.

1 Like