1. Re: Random number generators
Reading the recent posts and the apparent misunderstandings they are laden
with, I just have the following question for RobC:
You can't expose the algorithm too widely, so don't, but you know what it does.
Let D(n,p) the distribution of the sequence "result" after running the
following:
constant N=n, P=p
sequence result result=repeat(0,P)
for i=1 to P do result[i]=rand(N) end for
In theory, D(n,p) should be flat for all n and p. In practice, that
can't really hold true. But D(n,p) is probably flat enough for some range of
{n,p}. Different tasks require flatness for different specific values of {n,p}.
Question is: for which range can D(n,p) be considered as flat? What kind of
first-order deviation do you get outside the former range, so that
renormalizing to flat distribution can be performed if need be?
btw you can always consider result as an integer (the sum of the
result[i]*N^(i-1) ), if this is more convenient.
TIA
CChris
