A review on the statistical significance of VAN predictions

A Special Issue (Geophys. Res. Lett. 23, 1996) was focused on the question whether the VAN predictions outperform random chance. The majority of the p articipants of this Debate was selected to be critics against VAN, but Varo tsos and co-workers accepted to participate and responded to the critical c omments. VAN critics, including Geller (1996), share the same "requirements "; which decrease the success rate by 50% and the alarm rate by a factor 4- 5. Furthermore, these "requirements", when applied to an ideally perfect ea rthquake prediction method (which successfully predicts all earthquakes abo ve a certain threshold, and does not issue any false alarm), lead to the fo llowing paradoxes: (a) the success rate is below 100%, (b) the alarm rate i s 23% only, and (c) the ideal precursors are "postseismic" signals. Recent statistical treatments by Hamada (1996) and by Aceves et al. (1996) coincid e to the conclusion that VAN predictions cannot be ascribed to chance. (C) 1999 Elsevier Science Ltd. All rights reserved.