Random Number Generator tests

The Random Number Generator was a true hardware random number generator based upon noise in a zener-diode. This noise is supposed to be due to quantum tunneling of electrons.

Before and after the morning and evening sessions long runs without a subject present were done. We tested the RNG results for zeroth,first, and, second order effects, that is bias on bit frequencies, for alterations ('10' or '01' diads) or for one skipping alternating (eg. '1x1' etc.)

The results of these checks were consistent with previous series of tests in which we found on a total of 101211635 samples that there were slight deviations from ideal behaviour. The probability for a '1' was 0.50037 rather than 0.50000 and the excess alternating over persisting bits was 0.153%. One skipping alternating did not show any detectable deviation from chance behaviour.

The simple bias effect is compensated for by balancing the number of runs for each target alternative. The second, alternation, effect may give rise to a slight reduction in empirical variance so that the "alternation z-scores" are a bit conservative assuming a theoretical variance rather than this empirical one.