Two techniques have been discussed in Data Points for testing if a single observation "with a value that does not appear to belong with the rest of the values in a data set" can be declared as an outlier. In the Data Points column, "Dealing with Outliers" (SN, Nov./Dec. 2008), the problem associated with one outlier masking another outlier in a single outlier test was mentioned, and a reference to the generalized extreme studentized deviate (GESD) was provided as a robust and comprehensive technique to effectively identify multiple outliers. This column provides a simple example of outlier masking and how to apply GESD to identify multiple outliers.
展开▼