PNNL

Abstract

We present multivariate seismic event identification methods that can be applied to a large number of highly correlated regional discriminants. The methods employ the ridge discrimination techniques first proposed by Smidt and McDonald (1976). Ridge discrimination was developed to address the problems associated with discrimination in high-dimension, colinear settings and is readily adaptable to linear, quadratic, and outlier identification rules. Ridge discrimination is a special case of regularized discrimination analysis (RDA) developed by Friedman (1989). RDA includes linear discrimination (LDA), quadratic discrimination (QDA), and Euclidean distance-based nearest mean discrimination in its parameterization.