I cannot remember which book says an equation should not be counted as a model. Can an equation be called a model? I have a formulation:
T = Tp + Tt + S/v
where T is the total time we take to travel a distance between points A and B in a community; Tp is the time taken to prepare the trip; Tt is the total time delay due to traffic friction; S is the total distance between the two points; v is the walking speed.
I see that T is a system, not just a variable. Besides, the variables on the right hand side are the components of T. These components seem not be "causal factors" that influence T, but they seem to be "disaggregated components of T". So, the equation can be written as:
System = component1 + component2 +...+ component4.
with all measured in minutes.
Is this equation a model in a static system? However, this equation is more like a definition than a causal explanatory system. I mean T is not 'caused' by Tt, for example, but it 'consists of' Tt,.., and v.
Equation and Model
I have converted such an equation into a dynamic model. You may want to develop it for your own before observing my model below, to see if you will come up with the same structure.
Here is my dynamic structure
Level: Distance travelled = INTEGRA(+walking speed, INITIAL DISTNACE)
Rate: walking speed=STEP(1, total delay time)*IF THEN ELSE(gap>0, NORMAL WALKING SPEED)
constants:
NORMAL WALKING SPEED = 30 m/min
INITIAL DISTANCE = 0 m
TOTAL DISTANCE = 600 m
Tp = 3 min
Tt = 2 min
auxiliary:
total delay time = Tp + Tt
gap = TOTAL DISTANCE - distance travelled
Finally, I perform Monte Carlo Simulation by varying all the parameters except total distance. The purpose is to see how many minutes I would take to travel such a distance. I found that it takes between 20-30 min, with 100% of simulation runs.
Which level of confidence (50 - 100%) should I use to present the result? Are there some notes on how to select that confidence level?
I was also wandering if you could comment on this modeling and analysis?
In fact, sensitivity analysis can be done using the static model shown in the first post, by setting the most pessimistic values for all the parameter on the right hand side of the equation. Which approach is better?
Here is my dynamic structure
Level: Distance travelled = INTEGRA(+walking speed, INITIAL DISTNACE)
Rate: walking speed=STEP(1, total delay time)*IF THEN ELSE(gap>0, NORMAL WALKING SPEED)
constants:
NORMAL WALKING SPEED = 30 m/min
INITIAL DISTANCE = 0 m
TOTAL DISTANCE = 600 m
Tp = 3 min
Tt = 2 min
auxiliary:
total delay time = Tp + Tt
gap = TOTAL DISTANCE - distance travelled
Finally, I perform Monte Carlo Simulation by varying all the parameters except total distance. The purpose is to see how many minutes I would take to travel such a distance. I found that it takes between 20-30 min, with 100% of simulation runs.
Which level of confidence (50 - 100%) should I use to present the result? Are there some notes on how to select that confidence level?
I was also wandering if you could comment on this modeling and analysis?
In fact, sensitivity analysis can be done using the static model shown in the first post, by setting the most pessimistic values for all the parameter on the right hand side of the equation. Which approach is better?
what is a model?
Hi
For me a model is a representation of something using specific material, that looks more or less like the something in question. The material can be mathematic or painting or anything else.
So Y = 2 * X is for me a model too.
But I do not think it has any importance.
I think you are too much concerned by theoretical questions.
Why transform a simple equation into a dynamic model written in Vensim, unless you know why.
You can too keep the simplicity of the equation and represent it in static model with one equation, where the time is a constant.
I think that the most important thing in modellins is to know why one does something, and avoid guessing.
Why are you trying to change the equation in a dynamic model?
The best is eventually to post a model about your questions.
But beware about making things more complicated than they are.
A question found in Pat Rivett's book about decision modelling. What is the shortest way to split a chocolate bar of 6 line by 4 rows to get it to its 24 pieces?
Regards.
JJ
For me a model is a representation of something using specific material, that looks more or less like the something in question. The material can be mathematic or painting or anything else.
So Y = 2 * X is for me a model too.
But I do not think it has any importance.
I think you are too much concerned by theoretical questions.
Why transform a simple equation into a dynamic model written in Vensim, unless you know why.
You can too keep the simplicity of the equation and represent it in static model with one equation, where the time is a constant.
I think that the most important thing in modellins is to know why one does something, and avoid guessing.
Why are you trying to change the equation in a dynamic model?
The best is eventually to post a model about your questions.
But beware about making things more complicated than they are.
A question found in Pat Rivett's book about decision modelling. What is the shortest way to split a chocolate bar of 6 line by 4 rows to get it to its 24 pieces?
Regards.
JJ
Model
Hello JJ,
Thank you for your help. As to the distinction between a model and an equation, it find it important in writing a research report. The word 'model' signifies something superior to the word 'equation' as people are more familair with equations than models. If I call my equation a model, it looks new to them.
In addition, equations may or may not simlify the real world, while models always do. If I can my equation a model, I am implicitly asserting that it is a simplified system.
I translated the simple equation to Vensim ones because I would like to conduct a Monte Carlo simulation, to produce a more complicate output. Sometimes journal referee prefers complex results to simple ones. The graph with confidence bound looks good. It looks as good as the P-value in statistical analysis, which can make the results more plausible to some extent.
I want to make it more complicate than it is for adcademic reasons. It is just like when modelers have to aggregate their models just to satisfy their audiences.
As to the model, I have attached it here for your comments. Thank you again.
Thank you for your help. As to the distinction between a model and an equation, it find it important in writing a research report. The word 'model' signifies something superior to the word 'equation' as people are more familair with equations than models. If I call my equation a model, it looks new to them.
In addition, equations may or may not simlify the real world, while models always do. If I can my equation a model, I am implicitly asserting that it is a simplified system.
I translated the simple equation to Vensim ones because I would like to conduct a Monte Carlo simulation, to produce a more complicate output. Sometimes journal referee prefers complex results to simple ones. The graph with confidence bound looks good. It looks as good as the P-value in statistical analysis, which can make the results more plausible to some extent.
I want to make it more complicate than it is for adcademic reasons. It is just like when modelers have to aggregate their models just to satisfy their audiences.
As to the model, I have attached it here for your comments. Thank you again.
equation versus model
Hi
It is a very complicated way to represent something, even if it is to bluff the audience. If somebody in the audience has some very limited notions of maths, he will wonder why you make such a simple problem so complicated.
On top of that you model does not show the total time spent.
See model joined that calculates the time taken.
To make things more complicated and to exploit the power of SD, you can make the time spent preparing depending on the total distance using a table function, the delay spent in traffic too, and make the speed decrease with the fatique that is proportionnal to the distance already covered and can imagine very intricate feed back loops that will impress the audience much more.
Regards.
JJ
It is a very complicated way to represent something, even if it is to bluff the audience. If somebody in the audience has some very limited notions of maths, he will wonder why you make such a simple problem so complicated.
On top of that you model does not show the total time spent.
See model joined that calculates the time taken.
To make things more complicated and to exploit the power of SD, you can make the time spent preparing depending on the total distance using a table function, the delay spent in traffic too, and make the speed decrease with the fatique that is proportionnal to the distance already covered and can imagine very intricate feed back loops that will impress the audience much more.
Regards.
JJ
- Attachments
-
- walking model_1.mdl
- (3.36 KiB) Downloaded 403 times
Revised model
What a creative structure! I very much like the way you represent the total travel time in the model. That might use a lot of SD experience.
The attachment below is a revised model. I added another factor for Vensim to tell us when the destination is reached. Please run the model. You will see a dialog box showing the error found exactly when the destination is reached.
Do you consider this practice good? As I could see, with that factor the model can be judged invalid, but it is the quickest way for us to observe the total travel time, as we do no longer need to click any levels to see what the total travel time is. Vensim will tell us automatically. To be sure, that error occurs beyond the time horizon of interest, the time interval [0, total travel time].
The attachment below is a revised model. I added another factor for Vensim to tell us when the destination is reached. Please run the model. You will see a dialog box showing the error found exactly when the destination is reached.
Do you consider this practice good? As I could see, with that factor the model can be judged invalid, but it is the quickest way for us to observe the total travel time, as we do no longer need to click any levels to see what the total travel time is. Vensim will tell us automatically. To be sure, that error occurs beyond the time horizon of interest, the time interval [0, total travel time].
Adding the loops really makes the model more complicate and more difficult to be documented within a limited 12-page manuscript too. So, I have to assume that the motivation time, the traffic delay, and the walking speed are simply constants. This is a reasonable assumption because we know that the total distance is short (600 m) and thus unlikely to influence the plausible ranges of these parameters. Perhaps to add the loops will impress the reader less, a counterintuitive behavior of the reader in respose to our modeling.
Your point, however, is to describe how I can make the model much more complex. Thank you very much indeed.
Your point, however, is to describe how I can make the model much more complex. Thank you very much indeed.
equation and model
Hi Monte
I do not think that Bob would like the way you use Vensim.
You choosed to represent a very simple problem that can be expressed simply , see static model joined, in a dynamic fashion. The result cannot be calculated before the model has reached its final time. So to get the result you have to look at results given at the final time.
There is a very big difference between you and me.
You are probably a new comer to SD, I have already been using SD for more than 5 years.
So you are still thinking that the utility of a model is proportionnal to its complexity. I believe the contrary.
Regards.
JJ
I do not think that Bob would like the way you use Vensim.
You choosed to represent a very simple problem that can be expressed simply , see static model joined, in a dynamic fashion. The result cannot be calculated before the model has reached its final time. So to get the result you have to look at results given at the final time.
There is a very big difference between you and me.
You are probably a new comer to SD, I have already been using SD for more than 5 years.
So you are still thinking that the utility of a model is proportionnal to its complexity. I believe the contrary.
Regards.
JJ
- Attachments
-
- walking model_3.mdl
- (2.22 KiB) Downloaded 427 times
Hello JJ,
Thanks a lot. I laughed at your static structure for it looks unlike all other Vensim models I have been met. Then, I became highly appreciated with it after doing a Monte Carlo simulation. It is a perfect structure, exactly consistent with the equation I considered.
I, too, don't think Bob like my way of using Vensim, but I could not find a better way until seeing the way you use it, the right way.
It is good to hear you have used SD for a long time. I am new to SD, and always welcome your comments or any ideas deemed good to a beginner like me.
As to Coyle's discussion. Homer and Oliver did really focus another point. Coyle's issue is the improper use of multiple multipliers, while Homer and Oliver's issue is the additional value of formal model. Thing happens as you stated earlier. Thank you for your insights and emphasis.
Best wishes,
MK
Thanks a lot. I laughed at your static structure for it looks unlike all other Vensim models I have been met. Then, I became highly appreciated with it after doing a Monte Carlo simulation. It is a perfect structure, exactly consistent with the equation I considered.
I, too, don't think Bob like my way of using Vensim, but I could not find a better way until seeing the way you use it, the right way.
It is good to hear you have used SD for a long time. I am new to SD, and always welcome your comments or any ideas deemed good to a beginner like me.
As to Coyle's discussion. Homer and Oliver did really focus another point. Coyle's issue is the improper use of multiple multipliers, while Homer and Oliver's issue is the additional value of formal model. Thing happens as you stated earlier. Thank you for your insights and emphasis.
Best wishes,
MK