Sensitivity2All

[geo_curious]

Hi,

I am running a Sensitivity2All. Each run I do is reflecting a floating point error in the S2A results. There does not appear to be any variable that I select that does not return a floating point error. The simulation will run fine, as is.

Does anyone have any advice on this for S2A?

[Administrator]

What does S2A mean?

If you have a FP error on a specific run, right click on the bar in the Sens2All output and you can run that particular combination of constants to see where the FP error is generated.

[tomfid]

Surely S2A = Sensitivity2all

[Administrator]

Surely S2A = Sensitivity2all

Thanks Tom, obvious really, confused me completely when I read it.

[geo_curious]

From the S2A screen, I chose to simulate 1 param +/- 10%.

Screenshot 2024-05-09 081454.png (165.52 KiB) Viewed 11868 times

The result shows a series of failures.

Screenshot 2024-05-09 081539.png (535.41 KiB) Viewed 11868 times

I do not actually see anything different in return - I am not sure where the results of the simulation run go from selecting within the S2A bar. When I try to rerun S2A after that, Vensim crashes and closes.

[geo_curious]

I have also tried creating a list of some selected parameters from the advanced tab. They all also returned floating point errors.

A uniform variation on a single parameter, which works fine with regular simulation runs, also does not work when attempting a basic sensitivity run.

[Administrator]

I think we'd need to see the model to get any further with this.

[tomfid]

The thread affinity warnings are benign. It just means the operating system declined to assign requested threads. Presumably this is Windows, where thread management is not as good as Linux. It won't affect results.

The crash might be related to the floating point errors in the model. Did you submit the crash report?

[tomfid]

You should be able to replicate the error manually. Just do a run with the first reported problem, manually changing additional cooling tariff kwh as shown.

When you choose "simulate 1 param +/- 10%" it creates ordinary runs that will appear in the dataset control, on plots, etc.

[geo_curious]

There is no floating point error if I reduce the tariff by 10%, manually. I will send the model.

[geo_curious]

An update - I did send the model on Friday by email to the Ventana info email address. I have been trying uniform and random normal sensitivity runs this morning and there is no change. Vensim does not return any differences for anything but the changed parameter. If I do it manually, it works fine. There is something wrong with Vensim DSS in this case.

[Administrator]

What email address did you send it to? Nothing has arrived at vensim@vensim.com.

[Administrator]

Also, are you on the latest Vensim?

[geo_curious]

I did send the model on Friday by email to the Ventana info email address.

Thank you for sending the other email address. I will resend now.

I am using the latest version, yes.