How to perform global sensitivity analysis on a set of data

I have six independent variables, each taking 11 discrete values 0, 0.1, 0.2, ….1, and many dependent variables (continuous). For simplicity, let’s say one dependent variable. So that is 6 + 1 columns and over a million rows. I would like to do global sensitivity analysis (variance based, Sobol indices) but I cannot find how. I think Matlab offers something similar within SimBiology.

Answer 1

[Star Strider]

This is new to me. The only reference I was able to find (other than the SimBiology functions) is this GitHub link Global sensitivity analysis . There are several others, including on from the University of Arizona that appear to be promising as well.

I found these with a DuckDuckGo search for ‘sobol indices matlab’.

.

[Miroslav Blagojevic]

Thank you for the quick answer. I think this could be useful. I have never used DuckDuckGo, which is probably why I have not seen this GitHub text.

It is strange that Matlab offers GSA (which is not a rare analysis) only for a small set of problems (SimBiology), but not for anything else.

[Star Strider]

As always, my pleasure!

SimBiology appears to be a world unto itself. Its differential equation solvers (at least for kinetics problems) apparently use the Gillespie algorithm (a sort of ‘first principles’ approach that I find difficult to code for my own use), the ability to easily fit kinetic differential equation systems to data (this is noit diffiicult in base MATLAB however still something of a challenge) and several other special functions. I generaly do not do the sort of research SimBiology is designed for, so I have no reason to buy it, however that does not decrease the general utility of many of its functions and my desire to be able to use them.

I feel your pain!

.

[Miroslav Blagojevic]

It's not that bad; I still like Matlab :)

Thank you again.

[Star Strider]

So do I!

As always, my pleasure!

.