www.modeler-forum.com

Hello,
I am a bit puzzled by a situation that I am finding (I am probably just plain stupid): with iModeler the more intermediate factors there are between factor A and target factor B, the smallest the effect of factor A. Eg: more rain on the upper part of watershed leads to more sediments in stream A, and if there are more sediments in stream A then the next stream will have more sediments, etc.. which leads to more sediments in a dam. In the insight matrix the factor that has the smallest impact on the dam is Rainfall, but it is the cause of all sediments! Could you tell me what I am doing wrong?
Another example would be a volcanic eruption which long cascade of effects could be that I have no electricity in my house. Again the least affecting factor of having no electricity in my house is the volcanic eruption..
Thanks
Grégoire

Another question I have.. Suppose a factor A has a bunch of negative influences on it (e.g. -100), an that factor A influences positively a factor B. Since the effect of A on B is Value of A*infuence(A->B), can the influence become negative after a few time steps (since the value of A may become negative)?

Good questions

Answering your first question:
A strong impact just translates into a strong impact. The aim is to compare impacts and the sum of impacts in comparison to other impacts. For your purpose, though, you do not want to compare impacts, but track one impact along the cause chain. Simply set the weight of the impact to 100 percent and then you will get the result you expect. Rain would then have the same impact as the sediments of the upper level.
However, the sediments shouldn't have less impact than the rain. If the sediments could be prevented rain wouldn't be a problem anymore, would it?

Your second question:
That's a limit to a qualitative model. A factor can be named "Happiness" and then it could mean more or less happiness, but in cannot become sadness. Sometimes it is useful to add an explicit second factor, so that e.g. stress on the job leads to less happiness in the short run while it could lead to sadness in the long run (which is indicated by the long term effect set in the connection's properties).

Does that help?

Happy modeling
Kai

_________________
Kai Neumann

Consideo

Thanks a lot Kai, this helps, indeed.. Seems that we have to be very careful when assigning weights in a causal chain!

In the same series of questions, I have the following: I wanted to model "synergies", i.e. that two factors that are synergetic have greater impact than individual factors. So I have e.g.:
A --(0.1)-> Target
B --(0.1)-> Target
A --(0.5)-> B
B --(0.5)-> A

Looking at the impact matrix, I find that the effect of A or B on the target is 17.5 (short term), 27.5 (medium term) and 37.5 (long term). I am not sure I understand how I can get these values, because the feedback loop on A or B does not include the (0.1) effect on the target, so how do I get it to influence A or B?...

Thanks in advance..
Grégoire

Here is the model with the answer to the question:

https://www.imodeler.info/ro?key=AnyLlV ... -HKLmVDpXg

_________________
Kai Neumann

Consideo

Thanks a lot Kai, this is extremely useful! I know that nobody cares for the math, but I check it when I find counter-intuitive results that I cannot explain to my colleagues. Please try now to reverse the sign of A->B and of B->A, you will find that despite the feedback loop, the impact of A and B in the insight matrix is 7.5 and stays like that in medium and long term.. I suppose that we should have a decreasing effect as time passes...

If I do the same math (based on your descriptive text in presenter), it does not add up:

Here are the indirect impacts that already show that for each, a and b, the impact is the direct 10 plus the indirect -2.5.
The -2.5 e.g. of a is -0.5*10 (via b) plus -0.5*(-0.5)*10 (via b back to a) = -5+2.5.

(until now, ok.)

Over time this impact decreases because of the negative feedback loop it forms. The algorithm runs two time steps between short and medium as well as between medium and long term. Thus an impulse from a running through the loop increases -0.5*(-0.5) resulting in 0.25 that is taken two times and impacts the impact of a (0.25+0.25)*10 as well as the impact of b (0.25+0.25)*10. Well, and both adds to 10.

(here, it does not work...)

Any hints?

Regards,
Grégoire

I am a visual thinker could you provide the model for this?
Would do you mean by "reverse"? If both connections are negative it is still a reinforcing feedback loop. Maybe the model helps me to understand.

_________________
Kai Neumann

Consideo

Here: https://www.imodeler.info/ro?key=A8xyVC ... TBDXYN4Vxg
Thanks!

For factor 2:
10 plus -2.5 = 7.5
The loop: 0.5*0.5*10 * 2 (=5 via factor 2) plus -0.5 * -0.5 * -10 * 2 (= -5 via factor 1) => 0

If you delete temporarily the connection between factor 1 and objective, look at the insight matrix, then undo this and delete the connection between factor 2 and objective and again look at the insight matrix ... you will see both, -5 and 5 as y-values.

It is a bit tricky but pure logic. Just an impulse and its effects.

_________________
Kai Neumann

Consideo