|
|
|
|
Subject: Cormas random numbers and sensitivity analysis
From: Christophe Le Page (lepage@cirad.fr)
Date: mer fév 20 2002 - 18:41:22 CET
Hi Niek,
Cormas is using the class RandFishmanMoore as random numbers generator. If you have a look at this class, you will see (#initialize
method) that the "seed" (an attribute named Ui) is chosen as the milliseconds of the clock. To force the sequence to start from the
same "seed", you should add in the #init method of your model this instruction:
RandFishmanMoore Ui: 666666.
Obviously here 666666 is an example, you could enter any other small integer. Then the seed will be forced to that position, each
time the same, and you should have the very same results for two runs of a "stochastic" model. You should first verify this ! Then
if it's the case, you can be quiet confident in performing a valuable sensitivity analysis by changing slightly a parameter's value
!
clp
From: <Niek.Koning@Alg.AAE.WAU.NL>
> To: <cormas-owner@hermes.cines.fr>
> Sent: Tuesday, February 12, 2002 8:28 PM
> Subject: Cormas specialization / random numbers
>
>
>
> Another question: when doing sensitivity analysis, how can I secure
> that each simulation run uses the same sequence of random numbers?
>
> Greetings,
> Niek Koning