Hypothesis-Testing Approach to Outlier Detection

摘要

In this report we develop a general robust near-optimal methodology for testingfor outliers to one or more populations. The methodology requires no distributional assumptions and allows imperfect data sets, i.e., data sets with missing observations. The data can be from one or several seismic stations or can be a mixture of seismic and nonseismic data and continuous and categorical data. Listings of the computer code required to implement the methodology are included. (AN).