Is it possible to scale outputs of Morris analysis?

[sitadrost]

I've also posted a more general version of this question on Cross Validated (stats.stackoverflow), here
Is it possible in SALib to scale the outputs (mu, mu_star, sigma) of SALib.analyze.morris, so that the sensitivity of different quantities of interest can be compared? I couldn't find anything about this in the docs, is there a way to achieve this?

[willu47]

Moret et al (2019) normalise mu by mu_max which allows them to compare the influence of parameters across multiple outputs.

[sitadrost]

Thanks for the suggestion, that's not exactly what I'm looking for though. If you normalise that way, each QoI gets a maximum mu_star of 1, so you can't see the difference between a very robust QoI and a very sensitive QoI.

In the paper referenced on Cross Validated, they scale the elementary effects for each QoI by the ratio of (standard deviation of input parameter)/(standard deviation of QoI), which sounds sensible enough, except that I'm not sure how this works out for input parameters that are not uniformly distributed in [0,1]