|
|
|
|
Subject: Fw: normal distribution
From: francois_bousquet (francois_bousquet@hotmail.com)
Date: sam déc 01 2001 - 05:04:07 CET
Hi Damien
Thank you for your contribution. For the normal distribution I recently
wrote the following method in response to an email from Jeljer. The zip
attached contains a Cormas model and the graphical test.
I realise that this is the same algorithm, I just multiplied by sigma and
added mu.
So we have consistent methods.
For Info : I programmed it from the Simulation reference book : Law, A. M.
and K. W.D. (1991). Simulation modeling and analysis, Mc Graw-Hill
international ed.
!Cormas class methodsFor: 'util_aleat'!
normal: mu sigma: s
"polar method, Law et Kelton"
| w u1 u2 v1 v2 y x1 x2 |
w := 2.
[w > 1] whileTrue:
[u1 := Cormas random.
u2 := Cormas random.
v1 := 2 * u1 - 1.
v2 := 2 * u2 - 1.
w := v1 squared + v2 squared].
y := (-2 * w ln / w) sqrt.
x1 := v1 * y.
x1 := mu + (s * x1).
^x1! !