Re: 32-bit random numbers
- Posted by "Igor Kachan" <kinz at peterlink.ru> Jul 13, 2004
- 576 views
Hello Ricardo, You wrote: >Date: Jul 13 5:29 >From: "Ricardo M. Forno" <rforno at uyuyuy.com> >Subject: RE: 32-bit random numbers >Hi, both Igor and Al. >Actually I didn't say anything about these formulas. >In an older post, I was referring to another formula >saying not that it wasn't flat, but that it didn't >generate all possible numbers. It was flat, however. >Regarding Al's formula and Igor's formula, it seems >to me they both produce flat distributions, but this >is only shallow thinking and I haven't yet tried >to prove their flatness. >Regards. I'm sorry, Ricardo, yes, it was in Al's thread Re: Let's try this ONE more time Your words are: ------ > > Hi, Al. > > You said: > > <snip> > > > 1. For two added rnd numbers the distribution > > should be the same. > > It's multiplication that changes the distribution. > > In fact, adding two or more random numbers > *changes* the distribution. > Old FORTRAN routines used the sum of 12 random > numbers (why exactly 12? Because this was > considered enough to get a good distribution) > to get a Gaussian normal distribution. > > Regards. Yes, the simplest composition of 2 added random numbers gives not flat, triangle, Simpson's distribution, 12 gives almost already Gaussian not flat distribution. So, I thought you are saying about *not flat* distributions at all. But actually you are saying only about *changes* in more common sense, yes? Sorry again, my hurry. Anyway, your words and your restraint shows your good understanding of the discussing question. Thanks. Regards, Igor Kachan kinz at peterlink.ru