Re: Fwd: [Help-gsl] random number distributions

[John D Lamb]

Joseph Wakeling wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> John D Lamb wrote:
>   
> > There is nothing to stop you using two random number generators, though
> > you probably only need one. If you use two random number generators with
> > the same seed and call them equally often then the results will be
> > strongly correlated. If you use just one they will be nearly
> > uncorrelated. So I use one unless there's a good reason to do otherwise.
> >     
>
> Is there any means of selecting two seeds so as to guard against this,
> or will any two different seeds do the job as well as can be?
>   
There's nothing in GSL. In practice, this kind of problem ought to be 
unusual. The kind of place it might come up is where you regularly 
gernerate a pair of numbers, one from a uniform distribution, the other 
from an exponential. If you use separate gsl_rng objects for these but 
use the same seed for both, then you'll get highly correlated random 
variates. I would use a single gsl_rng but you could use two gsl_rng 
objects with different seeds as long as you weren't worried about nth 
order autocorrletaion effects where n is probably very large. Using 
different methods of generating random numbers would also work.

JDL