Am 18.04.2011 um 11:25 schrieb Heiko Bauke: > Hi, > > On Mon, 18 Apr 2011 10:16:50 +0200 > Toralf Niebuhr <[email protected]> wrote: > >> Question still is. If I have different random variables, should I use >> the same random number generator or should I have a new one for every >> random variable. > > if the PRNG is sufficiently random there is no benefit from using a new > PRNG for for every random variable. In fact, using many streams from the > same PRNG with different seeds may (depending on the seeding procedure > and the PRNG) introduce correlations. > > > Heiko
Sometimes, when doing process simulations, it is convenient that independent processes don't change their behavior, just because another process draws more random numbers (for example because of an implementation change). What one wants is, that every process only gets a different scenario, if one changes the initial seed (the generator seed). Which seeding procedure should I use if I want different independent streams? Toralf Niebuhr _______________________________________________ Help-gsl mailing list [email protected] http://lists.gnu.org/mailman/listinfo/help-gsl
