This article focuses on assessing discrimination performance for regional seismic events using a multivariate statistical method for outlier detection. The approach is motivated by the lack of seismic calibration data for underground nuclear explosions in most regions and the difficulty associated with transporting regional discriminants. The procedure is fully automated to flag events warranting special attention. It also allows control of the false alarm rate and a natural way to rank events. Regional data sets, consisting of nuclear explosions, mining blasts, and earthquakes, recorded by the ARCESS and GERESS arrays in Norway and Germany, respectively, station WMQ in China, and stations KNB and MNV in the western United States, are presented and compared. The outlier-test procedure is applied in each region, using Pn/Lg, Pn/Sn, and Pg/Lg in several frequency bands from 3 to 8 Hz. Identification and false alarm rates are estimated for each region using a standard set of discriminants, provided measurements were available for the events. At 0.01 significance level, between 87% and 100% of the nuclear explosions and quarry blasts are identified as outliers of the respective earthquake groups in the various regions. Overall, 209 of 229 (91%) explosions were identified as outliers, and there were only two false alarms out of 143 earthquakes (1.4%), slightly higher than the target rate of 1%. These results were obtained for diverse geological regions, for a wide range of epicentral distances and magnitudes, and for single stations and arrays.