hi.
how to validate the model in sd?
Thank you.
model validation
Hi
The notion of validity is partly subjective.
Tho only objective thing is the purpose if it is well defined.
One model will be considered as valid relatively to a purpose by someone, not valid by another person and half valid by someone else. So the way you will have to validate your model will depend closely on whom you will show your model to.
Regards.
JJ
[Edited on 4-11-2010 by LAUJJL]
The notion of validity is partly subjective.
Tho only objective thing is the purpose if it is well defined.
One model will be considered as valid relatively to a purpose by someone, not valid by another person and half valid by someone else. So the way you will have to validate your model will depend closely on whom you will show your model to.
Regards.
JJ
[Edited on 4-11-2010 by LAUJJL]
I do not pretend to represent any official point of view. It is a personal one.
For instance, I reviewed the last Seoul conference and looked at the models in Vensim that you could download.
All the models were coming from academics. There were 16 models downloadable.
From the 16 models, 14 were not valid from my point of view, because they either had no units or had unit errors not corrected.
Only one of the 16 models had a mass balance check and it was a model where all the variables had unit dimensions but where there were a lot of unit errors.
One of the two models without any unit error, could afford not to have a mass balance check, the model being simple but to my opinion not the other one.
So in the final, from the 16 models only one was eventually valid. After sudying it further on, I found that it was very reasonnably valid relative to its purpose.
If you ask the authors of the 14 unvalid model from my point of view, I am sure that they consider them as valid, especially if they published them as papers at the SD conference. If you ask the people who reviewed the papers, they will tell you that they are valid too.
As a remark, models without any unit error is considered by academics to be one of the first priority to achieve when making a model and it should always be achieved whatever the circomstances.
Of course no unit errors does not mean that the model is valid at all, but it is a first necessity.
Regards.
JJ
[Edited on 4-11-2010 by LAUJJL]
[Edited on 4-11-2010 by LAUJJL]
For instance, I reviewed the last Seoul conference and looked at the models in Vensim that you could download.
All the models were coming from academics. There were 16 models downloadable.
From the 16 models, 14 were not valid from my point of view, because they either had no units or had unit errors not corrected.
Only one of the 16 models had a mass balance check and it was a model where all the variables had unit dimensions but where there were a lot of unit errors.
One of the two models without any unit error, could afford not to have a mass balance check, the model being simple but to my opinion not the other one.
So in the final, from the 16 models only one was eventually valid. After sudying it further on, I found that it was very reasonnably valid relative to its purpose.
If you ask the authors of the 14 unvalid model from my point of view, I am sure that they consider them as valid, especially if they published them as papers at the SD conference. If you ask the people who reviewed the papers, they will tell you that they are valid too.
As a remark, models without any unit error is considered by academics to be one of the first priority to achieve when making a model and it should always be achieved whatever the circomstances.
Of course no unit errors does not mean that the model is valid at all, but it is a first necessity.
Regards.
JJ
[Edited on 4-11-2010 by LAUJJL]
[Edited on 4-11-2010 by LAUJJL]
Nadia, have a quick look on this paper it will give you a great idea about SD validation methods
http://ieeexplore.ieee.org/stamp/stamp. ... r=05480403
http://ieeexplore.ieee.org/stamp/stamp. ... r=05480403