Jump to content

Sensitivity Analysis


Recommended Posts

Hello, I was wondering if someone can help me with doing sensitivity analysis in Simio? So I know of the input parameters tab where I can see how much does the Objective function change by changing the time input parameters by 1 minute. But, I'm looking to see how much does a variation in a property in the model, for instance changing mis-identification rate, on the DIDO time and objective function. My thought process was just to run different scenarios and take the difference between the two and that would tell me how much it has a change. Yet, my professor suggested the following and I was still confused about how to do it: 

"So you can ESTIMATE a derivative by just COMPUTING  (g(x + h) – g(x))/h for a small h. This is called “finite differences.”

Thus, you can estimate sensitivities by running simulation experiments at different settings. In your case, if p is the nominal probability of misidentification, you can run an experiment at p + h also, then estimate the sensitivity by taking differences.

Simio’s internal sensitivity analysis is able to estimate derivatives using a trick that works in some, but not all cases, without the need for an additional run. But when it does not apply you need a different method." 

Link to comment
Share on other sites

The best way to do a sensivity analysis is to make "designed experiments", run your simulation according to this desing one by one and record the results of them and then analyse the results by using "statistical methods". Thus, I suggest you to investigate "experimental design" or in other name "design of experiments" concept. At first, you can get confused with this topic thus you can ask for help from an experienced expert in this area. The above method is called "perturbation analysis" and as far as I know it is not the best way or at least a common method for doing sensivity analysis. 

Link to comment
Share on other sites

×
×
  • Create New...