A FEW CONSIDERATIONS FOR ASCRIBING STATISTICAL SIGNIFICANCE TO EARTHQUAKE PREDICTIONS

There are many potential pitfalls in assigning statistical significanc e to ''successful'' earthquake predictions, including sensitivity to t he stochastic model for earthquake occurrence, dependence among the pa rameters used to specify an earthquake, and the stochastic alternative model for the predictions. There is no simple resolution to these dif ficulties, but it is clear that assuming earthquake location and size variables are independent can lead to the erroneous impression that pr edictions are statistically significant when they are not, and that as suming that main shocks follow a Poisson process in time (and space) c an make chance predictions appear statistically significant, and can m ake sound predictions appear statistically insignificant. It is possib le to avoid assuming any probability distribution for the occurrence o f earthquakes, treating the seismicity history as fixed (conditioning on the observed seismicity) and testing the performance of the predict ion algorithm against various random predictors; this ''conditional'' approach seems preferable.