Hi,
Are there tools in Vensim that analyses
sensitivity analysis using either SWAT-CUP (SWAT calibration uncertainty procedure)
and SUFI-2 (Sequential uncertainty fitting ver 2)
or include "P factor" and "R factor" ?
R factors is the ration of the average 95PPu band and the SD of measured variables
P factor is the fraction of measured data bracketing by 95PPU band.
95PPU (95% predictive uncertainty )
sensitivity analysis
-
- Super Administrator
- Posts: 4573
- Joined: Wed Mar 05, 2003 3:10 am
Re: sensitivity analysis
I've never heard of SWAT-CUP or SUFI-2. Tom might have though (I'm sure he'll be online later). But the words "calibration" and "fitting" suggest these are optimization procedures, not sensitivity.
Can you calculate these R and P factors in your model and use them in your analysis?
Can you calculate these R and P factors in your model and use them in your analysis?
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: sensitivity analysis
I hadn't either, but I assume it refers to this: https://swat.tamu.edu/media/114860/user ... watcup.pdf I'd probably have to go back to the source papers to find out what's really going on, but this discusses parameter ranges as if the problem were univariate.
A key benefit of MCMC is that it generates a sample representing the joint, multivariate distribution of parameters. That makes talking about univariate bounds (especially uniform univariate bounds) on parameters a little unnatural.
The corresponding procedure in Vensim would be to use MCMC to generate a sample of points, then use the sample to generate sensitivity runs using the sample as file input.
A key benefit of MCMC is that it generates a sample representing the joint, multivariate distribution of parameters. That makes talking about univariate bounds (especially uniform univariate bounds) on parameters a little unnatural.
The corresponding procedure in Vensim would be to use MCMC to generate a sample of points, then use the sample to generate sensitivity runs using the sample as file input.
/*
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: sensitivity analysis
Hi Tom,
I just exported sensitivity analysis results and I got columns with a prefix like "T1" until "T24".
What does it mean?
Which ones are low percentile, median and high percentile?
I just exported sensitivity analysis results and I got columns with a prefix like "T1" until "T24".
What does it mean?
Which ones are low percentile, median and high percentile?
Re: sensitivity analysis
The T values are times, which you get when you choose to modify names by time http://vensim.com/documentation/ref_dat ... _sensi.htm
Percentiles are only available in command scripts: http://vensim.com/documentation/sens2file.htm
Percentiles are only available in command scripts: http://vensim.com/documentation/sens2file.htm
/*
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
*/