[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
## Re: [Help-gsl] random number generators

**From**: |
X Statistics |

**Subject**: |
Re: [Help-gsl] random number generators |

**Date**: |
Mon, 18 Apr 2011 09:53:33 -0300 |

>
>* 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 I should use a new one. How do I obtain the deterministic*
>* sequence of initial seeds such that the different generators produce good*
>* independent sequences?*
>
You do not need different RNGs for differing random variables. For example,
you can generate from a normal distribution and from a gamma distribution
using the same RNG.
For instance, people working with Bayesian statistics run a Markov chain
Monte Carlo algorithm, such as the Gibbs sampler, with a single RNG (eg.
gsl_rng_mt19937). In that algorithm, you may sample from different
conditional distributions (normal, gamma, beta, multivariate normal,
Student-t, Wishart, etc) using the same RNG.
That being said, if you use GSL, then initialize your RNG and use it in your
whole program. (The same happens in R and Matlab.) Thus, you only need to
save one RNG SEED.
Ralph.