The results produced by a theoretical randomness test are definitive, but the interpretation of those results is up to the simulation developer. For example, a serial correlation test may indicate that the successive outcomes of a uniform-distribution RNG have a 0.000132 autocorrelation, which is low, but only the simulation developer can say if it's low enough.
This page last modified on 2 March 2005.