Dependence of Anomalous REG Performance on Run Length

In random event generator (REG) experiments yielding anomalous results, any evolution of effect sizes over extended runs of data collection could, in principle, give some insight into the fundamental mechanism of the binary probability distortions. Retrospective examination of large individual and collective databases acquired over many years of previous study proves only marginally capable of distinguishing any such functional dependence on run length from a constant effect size model, which does little to narrow the phenomenological possibilities. However, an unanticipated ancillary feature of this data treatment is the emergence of a suggestively small blockwise variance in the anomalous effect sizes, possibly indicative of sequential correlations in the data streams that are not present in the baseline data. If real, such correlations would imply that the mechanism of the anomalous effect is more complicated than a simple change of elementary binary probabilities, although the set of possible models is so large that no specific inference can be drawn at this time.