Hi everybody
pliiiisss look at the attached model... I need to understand why the two gaussian curves (one obtained with a mathematical formula the other with a Vensim model) show such different values at their maximum (one is correctly at 1 the other at 0.008).
Best regards
Lorenzo
The expression for the Gaussian is incomplete. The full pdf is:
1/SQRT(2*pi)/sigma*EXP( (x-mu)^2/2/sigma^2 )
That partially reconciles the two. They still differ, but I don't understand the rationale for the DIFF version, so I can't really comment on that, except to point out that it doesn't work for sigma=1.
tomfid wrote:The expression for the Gaussian is incomplete. The full pdf is:
1/SQRT(2*pi)/sigma*EXP( (x-mu)^2/2/sigma^2 )
That partially reconciles the two. They still differ, but I don't understand the rationale for the DIFF version, so I can't really comment on that, except to point out that it doesn't work for sigma=1.
Hi
thank you a lot.... a really stupid mistake.... the rationale of the Vensim model is to generate the Gaussian dynamically and non through a formula....
Lorenzo
One way to generate an approximate Gaussian dynamically is to use a delay chain with a lot of levels. The SMOOTHN or DELAYN functions are convenient for this. A single level of a smooth or 1st order delay is an Exponential or Erlang(1) distribution. As you reach a large number, it converges to Gaussian. The output of a SMOOTH3 already looks pretty close if you feed it a PULSE input.
tomfid wrote:One way to generate an approximate Gaussian dynamically is to use a delay chain with a lot of levels. The SMOOTHN or DELAYN functions are convenient for this. A single level of a smooth or 1st order delay is an Exponential or Erlang(1) distribution. As you reach a large number, it converges to Gaussian. The output of a SMOOTH3 already looks pretty close if you feed it a PULSE input.
Hi
thank you a lot for this precious suggestion...
Best regards
Lorrenzo
