Nonparametric methods in studies of the seismic regime

Abstract

The application of nonparametric statistical methods to the estimation of some characteristics of the seismicity regime is considered. Emphasis is placed on the behavior of the distribution tail (i.e., the distribution of the strongest events). The methods described do not use any assumptions concerning the distribution function. Confidence intervals are derived for the magnitude distribution function and for the Poisson intensity of the flow of seismic events. The probability that a previously recorded maximum magnitude will be exceeded during some future time interval T and the confidence interval of this probability are estimated. The distribution of the time to the nearest event exceeding the last maximum (to the nearest record) is derived. The nonparametric approach is most effective if the type of empirical data parameterization is unknown or there are grounds for doubting its adequacy.