Dear community menbers,
I'm running some experiment with my model. I run 100 observations per each experiments and I have different sources with an own drain.
I'm setting the statNumOut as output value per each drain.
All of them are ok, one is giving me some problems. It's like the source procude entities only for the first observation. Here is a picture.
Attached you can find also my model for reference.
Do you have any advice?
Essentially there is nothing wrong with your model. The results you are getting are to be expected.
What is happening is two things:
1) The init method in HamburgerSource will only ever produce 1 mu every time. The other sources all can produce different amount of MUs every time the code is called. This means that naturally they will produce different numbers of Mus on every run.
2) The paralogistic distriubtion which sets when the code is called, produces a very short range of times. Hence when you are running the model over just 2 hours it will produce 11 Mus every time because of point 1 and these short times.
I did run the experiment over 10 days and there was some variation but it was very small (min 1254, Max 1256). This shows that there is variation but requires a longer amount of time to get it.
Does that make sense?
It seems that i found the issue.
I typed the z_paralogistic in 2 ways:
1. z_paraLogistic(1, 2*60, 12*60)
2. z_paraLogistic(1, 2, 12)*60
when I type it with mode 1, it gives me the strange result I showed before, when I use mode 2, the results are the expected.
Do anyone knows why? I mean, it should be the same, doesn't it?
Been out for a couple of days but going back through my statistics notes from a few years ago I have got an idea of what it would be different
If your distribution was for example a Normal distribution (where the parameters are mean and standard deviation (SD) ) then you would be right in your assertion that the two ways are the same.
However you have chosen a distribution where the parameters are setting the shape of the distribution and not the mean and SD (Beta, Gamma etc are like this as well).
Therefore the two distributions you typed will have vastly different shapes and hence means and SD.
I can't tell you the exact mathematical reason but I know you can't change the shape by a constant and expect a similar result. This level of stats is a bit beyond me.
I have attached an example which shows 100 results of your distribution in a tablefile next too each other. I have also commented out the similar result but with a normal distribution to show how they are same. Just run the method and it will generate the results.
Here is a mathematical explanation:
The paralogistic function is calculated by the following formula:
z_paralogistic(alpha, theta) = theta * pow(pow(1-z, -1/alpha) - 1, 1/alpha), where z is a random number between 0 and 1.
==> z_paralogistic(alpha, theta) * 10 = 10 * theta * pow(pow(1-z, -1/alpha) - 1, 1/alpha) = z_paralogistic(alpha, theta*10)