Global sensitivity sensitivity/perturbation analysis

Question

Hi all

I am trying to do a global sensitivity analysis/perturbation analysis in OpenLCA, where all parameters need to have their values individually changed 10% to see the impact on the final result. However, this is very tedious work, so I was wondering if it possible to do automatically in OpenLCA in some way? Possibly using Python?

Comments

I agree, An O-A-T feature could be a nice contribution, where while making the model, the practitioner could "flag" parameters for the "O-A-T sensitivity analysis" feature, and in the feature panel later, the wanted input change could be chosen globally or edited per parameter.

The output would be a table of relative and absolut output change.

Would be nice. Not need though.

Answer 1

Hi, interesting question, this would work in Python but another, maybe simpler option could be to make a Monte Carlo simulation, if you specify the +/- 10% as boundaries of a uniform distribution, for these parameters. Only point would be that if you are using as generic database that has already uncertainty information, this could "disturb" your analysis.

Comments

How would this be possible in Python? Do you perhaps have a guide somewhere? Because I fail to see how it would be possible with either Python or using Monte Carlo simulation, as it is a One-At-A-Time (OAT) Perturbation analysis, i.e. only one value should be changed at a time.