It's fairly simple to perform if you use the morris test function in the sensitivity package of R. It supports decoupling which splits the sensitivity analysis into two phases: 1) creation of a design (you just need to supply information on the number of parameters and their ranges and the function returns the design); 2) running the model at the design points and passing the output back to the function. Your model can be executed in different software, you just need to store the output corresponding to the design in a vector and read it into R and pass it to the Morris Test object. I've probably made that sound more complex than it is. It's only a few lines of code.

If you use R I'm happy to provide an example model and code that demonstrates this process. Alternatively Matlab supports a variety of techniques for global sensitivity analysis.

Can you please provide an example?

#Create morris test design

#factors is the number of varying parameters (25) in the model

#binf is a vector of lower bounds for input parameters

#bsuf is a vector of upper bounds for input parameters

#r, levels and grid.jump are morris test options

Morris

Thank you very much.

Kevin McNally Hello, thank for your answer! Can you explain the last 4 lines code more specific?(Is tell a function and how can we get the output?)

Hello! I cannot seem to find an example online of a Morris sensitivity test in R, where y is modelled externally.

Does anyone have an example code of how outputs(Y) varied with inputs (X)?

Is this doable with the Morris method?

Say we have our 100 inputs x1,x2,x3,x4,...,x8 which cause our 100 outputs y1,y2,y3. And i have two dataframes modelX[100,8] and modelY[100,3].

Is there a way of doing morris(design=modelX, model=modelY, etc)?

Many thanks in advance!

I'm not sure I fully understand your question Melanie Jans Singh . Are you asking how to read in a vector of output for use with the Morris Test object in R?