Random Number Generating in R

Things under legendu.net/outdated are outdated technologies that the author does not plan to update any more. Please look for better alternatives.

1. Functions for sampling random numbers from distributions share a same "basic" random number generator (RNG). If one set a seed for the "basic" RNG in use, it affects all functions for generating observations from distributions. The kind of "basic" RNG can be queried and set by RNGkind. The default RNG in R is Mersenne-Twister.

2. When doing a big simulation, some people like to split the simulation into smart parts and run each part on a different machine. Theorectically speaking, this can cause problems, because random numbers generated on different machines might not come from disjoint parts of a same seed (or even not a same kind of random number generator). Parallel computing is an alternative to this approach.