Testing a Normal Sample for Multiple Outliers.
Abstract
Test statistics G and M for two outliers due to Murphy and Grubbs are compared with a test statistic R based on applying a single-outlier test to the sample with the most extreme observation removed. It is shown that G is generally most effective. The G and R procedures are shown to be equivalent when the number of outliers is unknown, and to give satisfactory performance for one or two outliers. An extension to the linear hypothesis is outlined. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Nov 01, 1976
- Accession Number
- ADA040327
Entities
People
- Douglas M. Hawkins
Organizations
- University of North Carolina at Chapel Hill