"Nobody ever gets credit" model

[Hamilton]

Hi.

I am trying to replicate the model depicted in the Reppening and Sterman's groundbreaking paper "Nobody ever gets credit for fixing problems that never happened" (http://web.mit.edu/nelsonr/www/Repennin ... _su01_.pdf).

I am new to system dynamics and Vensim, so of course I am having problems with the model. I would like to be able to replicate both kinds of behaviors depicted in the paper.

What am I doing wrong? I need help with the equations.

See the model attached.

Thanks in advance.

Hamilton.

[LAUJJL]

Which problems?

First I would correct the unit errors.

Another question.

The paper kindly exposes succesively more complex models, starting with very simple structures.
Did you take the opportunity to use the facility to build all the intermediate models to ease the understanding and to validate the written explanations given by the authors before pretending that the paper is so groudbreaking?

Personnally I would build each intermediate model, study it extensively and would go to the next one only after having totally understood the previous one.

Regards.

JJ

[tomfid]

It seems like the paper doesn't fully document the models and experiments, so it's hard to know what's missing.

Fixing units is a good start, but I'd email the authors and see if the model, or at least an equation listing, is available.

[LAUJJL]

I have looked again at the causal loop diagram of the model in the paper and I find it rather strange and difficult to transform in a plain model.
In particular there is no policy to implement in this model which is about what to favour, short term or long term.

I am modifying the structure of the model, to make it easier to model and that has explicit policy embedded.

i will join it when finished.

Regards.

JJ

[LAUJJL]

HI

It is a rather tricky problem, mainly because it is possible to modify the performance by increasing the worktime or/and the capacity and increasing the capacity is delayed and consumes time not available anymore to proper worktime.

I join a model that solves apparently the problem without guaranty. There are no systematic reality checks and I have enough spent time on it, especially as I do not know if the subject is interesting anybody anymore.

I join a packaged model that includes a run that starts in quasi stability, a second one where the desired performance jumps from 2000 to 3000 at time 50 and another one where the desired performance drops from 2000 to 1000 at time 50. The behavior looks correct.

Anybody more interested in the subject can post any question.

Of course the model is rather simplistic. There is no limitation to the capacity, and the possibility to increase it depends linearly from the time spent, which is not realistic, especially when one nears the limitation. One can too improve the model by adding the possibility to increase the total time available (the number of workers) the same way that it is done with the capacity, etc..

Regards.
JJ

[mike]

Thanks for posting this question/challenge/issue Hamilton.
Thank you JJ for your insightful comments.
Thank you Tom for a great quote :
"It seems like the paper doesn't fully document the models and experiments, so it's hard to know what's missing."
How many times has that happened to us all.
The challenge of trying to be SDers.