Can you please tell me if I want the output as a percentage change, that is If I am varying my input parameter with a 1% change then by how much will my output change using the any of the analysis method that is Sobol or morris?
Hello Dear, hope you are doing well. Reading your question, to me that sounds more to a local sensitivity analysis (LSA) related problem, which is even much simpler to run. In the case of Sobol or Morris methods, one usually is dealing with Global Sensitivity Analysis (GSA), and the investigation relies mainly on the contributions of model input uncertainty to model output uncertainty. In your case you can use Matrix Perturbation theory (The Computational Structure of Life Cycle Assessment, by Heijungs and Sangwon), or even simpler, you can run OAT One-factor-at-a-time by choosing one input, varying it by 1% while the remaing inputs are constant, run the model and see the effect of this 1% on the model output interested. Than repeat the procedure for the remaing parameters. I hope it helps. If you want to increase the reliability of your model, then move forward to Morris and Sobol, and for this you need an statistical analysis of the input because it is necessary the ranges (standard deviation) to see the variability of the input.
Will, this is fantastic. Thanks for sharing this great work
Great talk! Thanks for sharing!
Thanks for sharing
Can you please tell me if I want the output as a percentage change, that is If I am varying my input parameter with a 1% change then by how much will my output change using the any of the analysis method that is Sobol or morris?
Hello Dear, hope you are doing well. Reading your question, to me that sounds more to a local sensitivity analysis (LSA) related problem, which is even much simpler to run. In the case of Sobol or Morris methods, one usually is dealing with Global Sensitivity Analysis (GSA), and the investigation relies mainly on the contributions of model input uncertainty to model output uncertainty. In your case you can use Matrix Perturbation theory (The Computational Structure of Life Cycle Assessment, by Heijungs and Sangwon), or even simpler, you can run OAT One-factor-at-a-time by choosing one input, varying it by 1% while the remaing inputs are constant, run the model and see the effect of this 1% on the model output interested. Than repeat the procedure for the remaing parameters. I hope it helps. If you want to increase the reliability of your model, then move forward to Morris and Sobol, and for this you need an statistical analysis of the input because it is necessary the ranges (standard deviation) to see the variability of the input.
Thank you, I will apply this
can i do this method in parallel ?
Down in front!!