random numbers, random bits

From: Dale M. Slone (dale@wente.llnl.gov)
Date: Tue Apr 20 1993 - 01:08:29 UTC


I have 2 question regarding random numbers:

1) What statistical methods are appropriate to determine if 2 (or more)
   subsequences of random numbers are not correlated with each other.
    When I use rand() or rand48() I typically use <r*r>~.33333 and
   <r(i)*r(i+1)>~.250 as quick checks to establish that the individual
   sequences are random.  This is for a massively parallel application.

2) "Numerical Recipes" has a routine to generate random bits.  My
    particular application could use random bit generation, but I
    would like to bias the mean, i.e. instead of <r>=.5, I would
    like to generate a sequence of bits such that <r>=.5+epsilon,
    where epsilon is a small number computed based on the problem
    size.
    Is there such a method, and how would I establish that the
    sequence is random?

Any help would be appreciated.
Thanx
dale@frostedflakes.llnl.gov


This archive was generated by hypermail 2.1.7 : Tue Oct 14 2003 - 21:44:12 UTC