Hi,
I am quite new in SD and I am trying to use the method for my PhD. I took over the model from a former colleague. After correcting some equations, I can see my model is off, as it had been manually calibrated for the first "wrong" model. I am now considering the calibration module to give it a try, as there are numerous parameters and a manual one is a waste of time I think. I have read on the forum, the help manual, but this is still not clear.
I defined a payoff, i.e. one of my variable compared to a measurement one (time serie). I just try a parameter that I know would influence the results. and ... I am not exactly sure if something happens. I do not see any changes. I do not see any *.out file. Where am I supposed to find the updated values, if existing?
Any extra documentations, files or simple models with a calibration phase working would be highly appreciated...
Regards,
Greg
calibration procedure / example
-
- Super Administrator
- Posts: 4573
- Joined: Wed Mar 05, 2003 3:10 am
Re: calibration procedure / example
When registering, you entered zero as your serial number. Please correct this so we know you are a licenced user.
Advice to posters seeking help (it really helps us to help you)
http://www.ventanasystems.co.uk/forum/v ... f=2&t=4391
Units are important!
http://www.bbc.co.uk/news/magazine-27509559
http://www.ventanasystems.co.uk/forum/v ... f=2&t=4391
Units are important!
http://www.bbc.co.uk/news/magazine-27509559
Re: calibration procedure / example
I think an initial manual calibration is NOT a waste of time; it's very helpful for developing some intuition about what matters and what is plausible. Normally, I set up a simple control panel showing the relevant model-data comparisons and parameters. Then I do the calibration in synthesim. This requires that the model be reasonably fast (at most a few seconds). How many is "numerous"?
/*
Advice to posters (it really helps us to help you)
http://www.ventanasystems.co.uk/forum/v ... f=2&t=4391
Blog: http://blog.metasd.com
Model library: http://models.metasd.com
Bookmarks: http://delicious.com/tomfid/SystemDynamics
*/
Advice to posters (it really helps us to help you)
http://www.ventanasystems.co.uk/forum/v ... f=2&t=4391
Blog: http://blog.metasd.com
Model library: http://models.metasd.com
Bookmarks: http://delicious.com/tomfid/SystemDynamics
*/
Re: calibration procedure / example
Hey Tomfid,
You are correct, manual calibration is a good thing for the reasons you mention. "Waste of time" was not exactly what I meant, but more tedious procedure. The automatic calibration helps to verify if you can get reasonable values of parameters and reproduce the dynamic of the system. In the meantime, I managed to make the calibration work which certainly is a good thing.
Is there a way to load the calibrated parameters into the model afterwards to keep working from there?
You are correct, manual calibration is a good thing for the reasons you mention. "Waste of time" was not exactly what I meant, but more tedious procedure. The automatic calibration helps to verify if you can get reasonable values of parameters and reproduce the dynamic of the system. In the meantime, I managed to make the calibration work which certainly is a good thing.
Is there a way to load the calibrated parameters into the model afterwards to keep working from there?
-
- Super Administrator
- Posts: 4573
- Joined: Wed Mar 05, 2003 3:10 am
Re: calibration procedure / example
The optimization will produce a ".out" file, this can be used as a standard CIN file in Vensim (Model->Simulate, select "Changes", and "Load changes from").
Advice to posters seeking help (it really helps us to help you)
http://www.ventanasystems.co.uk/forum/v ... f=2&t=4391
Units are important!
http://www.bbc.co.uk/news/magazine-27509559
http://www.ventanasystems.co.uk/forum/v ... f=2&t=4391
Units are important!
http://www.bbc.co.uk/news/magazine-27509559