(Re-)Structuration of System Dynamics
Posted: Tue Jun 08, 2004 3:37 pm
Hi every one
I have consulted the list of papers presented at the next SD Conference in Oxford.
One has interested me very much at first it is:
>From Reichel, André Number 266 (Re-)Structuration of System Dynamics,
The first part is very interesting for me, summarizing relatively well the difficulties I encounter in trying to model my business problems; difficulty to go from qualitative to quantitative:
For example I have built a qualitative influence diagram that describes very well my problem and which is very readable, understandable and convincing even if it has about 60 variables and about 24 loops something important for my level in SD. But as soon as I try to change it into a quantitative model, everything is going bad. The realism in particular disappears because of the impossibility to determine realistic equations and realistic links between some model variables and reality.
I started back with a one or two loop model (a reinforcing loop with its associated balancing one), built the quantitative model, run it but the problem was the same only simplified:
a difficulty to measure the level and the kind of influence.
So I never developed a quantitative model, trying to get the best from the qualitative one.
I would like to find some intermediate step between full qualitative and full quantitative to
avoid the too high step going from one to the other.
It would then be possible to stay realistic and having some real data to compare to the reality.
In the first part of his paper André Reichel, says that the essence of the behaviour lies in the structure of the model, and that it is possible to draw conclusions only from its structure, avoiding the burden and imprecision (some times) of quantification.
I was of course interested and willing to see how it was possible to do that.
If the rest of the article, there is no proof that the method he proposes may avoid the necessity
of measuring things.
But I would like to clarify one thing:
Is structure the main explanation of the behaviour of a model?
In models of infection diseases, pages 300 to 323 in Business dynamic (to my opinion one of the best demonstration of the power of SD) a small variation of any parameters changes an infection into an epidemic by changing the influence of one loop on another.
For me it proves clearly that structure is not the only responsible of the general model behaviour. For me a model is a structure with its equations. Both are needed.
In an infectious disease the simple structure of the model without its equations, will not tell you anything about its possibility to degenerate into an epidemic.
So I cannot understand how it is possible to build a model by staying qualitative.
Has anybody an idea about this problem?
It is not the first time I read that structure governs behaviour.
Some days ago Geoff Coyle argued:
Some people show a slide of the stock/flow diagram and call it the model. It
isn't, it's a picture of the model and the real model is the equations. As
Jay pointed out, very many years ago, the defence of the model lies in the
defence of its details, that is in the equations, not in the behaviour.
About Geoff Coyle mail; he writes about Professor Pat Rivett's outstanding 'The Craft of Decision Modelling', Wiley 1994.
I could not find the book, nor on Amazon site nor on Wiley's.
Regards to everybody.
J.J. Laublé, Allocar rent a car company
Strasbourg France JEAN-JACQUES.LAUBLE@WANADOO.FR
I have consulted the list of papers presented at the next SD Conference in Oxford.
One has interested me very much at first it is:
>From Reichel, André Number 266 (Re-)Structuration of System Dynamics,
The first part is very interesting for me, summarizing relatively well the difficulties I encounter in trying to model my business problems; difficulty to go from qualitative to quantitative:
For example I have built a qualitative influence diagram that describes very well my problem and which is very readable, understandable and convincing even if it has about 60 variables and about 24 loops something important for my level in SD. But as soon as I try to change it into a quantitative model, everything is going bad. The realism in particular disappears because of the impossibility to determine realistic equations and realistic links between some model variables and reality.
I started back with a one or two loop model (a reinforcing loop with its associated balancing one), built the quantitative model, run it but the problem was the same only simplified:
a difficulty to measure the level and the kind of influence.
So I never developed a quantitative model, trying to get the best from the qualitative one.
I would like to find some intermediate step between full qualitative and full quantitative to
avoid the too high step going from one to the other.
It would then be possible to stay realistic and having some real data to compare to the reality.
In the first part of his paper André Reichel, says that the essence of the behaviour lies in the structure of the model, and that it is possible to draw conclusions only from its structure, avoiding the burden and imprecision (some times) of quantification.
I was of course interested and willing to see how it was possible to do that.
If the rest of the article, there is no proof that the method he proposes may avoid the necessity
of measuring things.
But I would like to clarify one thing:
Is structure the main explanation of the behaviour of a model?
In models of infection diseases, pages 300 to 323 in Business dynamic (to my opinion one of the best demonstration of the power of SD) a small variation of any parameters changes an infection into an epidemic by changing the influence of one loop on another.
For me it proves clearly that structure is not the only responsible of the general model behaviour. For me a model is a structure with its equations. Both are needed.
In an infectious disease the simple structure of the model without its equations, will not tell you anything about its possibility to degenerate into an epidemic.
So I cannot understand how it is possible to build a model by staying qualitative.
Has anybody an idea about this problem?
It is not the first time I read that structure governs behaviour.
Some days ago Geoff Coyle argued:
Some people show a slide of the stock/flow diagram and call it the model. It
isn't, it's a picture of the model and the real model is the equations. As
Jay pointed out, very many years ago, the defence of the model lies in the
defence of its details, that is in the equations, not in the behaviour.
About Geoff Coyle mail; he writes about Professor Pat Rivett's outstanding 'The Craft of Decision Modelling', Wiley 1994.
I could not find the book, nor on Amazon site nor on Wiley's.
Regards to everybody.
J.J. Laublé, Allocar rent a car company
Strasbourg France JEAN-JACQUES.LAUBLE@WANADOO.FR