Know Your Outliers
Using a new ASTM International standard will make it easier to simultaneously identify multiple outliers in a data set. The new standard is D7915, Practice for Application of Generalized Extreme Studentized Deviate (GESD) Technique to Simultaneously Identify Multiple Outliers in a Data Set.
"The impetus for developing D7915 was to provide a method for identifying outliers that is robust when multiple outliers are present in the data set," says ASTM International member Alex Lau, consultant, TCL-Consulting. "The practical application of the standard is for prescreening of interlaboratory study data, where there is a large volume of raw data to be screened for outliers."
Lau notes that traditional techniques require a user to decide in advance on exactly how many outliers to test for, as well as whether these outliers are excessively large or excessively small relative to the majority of the data set. However, the methodology described in D7915 only requires an approximation of total number of outliers to detect for.
In addition, the D7915 technique is well suited for automation, such as use in Excel, which is not the case with other classic outlier methods.
Primary users of this standard will be those involved in pre-screening ILS data. IN addition, this standard is well-suited for manufacturers that need to perform exploratory process capability analysis. Personnel involved in regulatory enforcement can also use D7915 to perform objective screening of field regulatory compliance data. Laboratories can use D7915 to properly assess test method capabilities by correct exclusion of outliers.
D7915 was developed by Coordinating Subcommittee D02.94 on Quality Assurance and Statistics, part of ASTM International Committee D02 on Petroleum Products, Liquid Fuels and Lubricants.
To purchase ASTM standards, visit www.astm.org and search by the standard designation, or contact ASTM Sales (tel +1.877.909.ASTM; sales@astm.org).
CONTACT Technical Information: Alex T. Lau, TCL Consulting • Whitby, Ontario, Canada • tel +1.905.665.7995 | ASTM Staff: Alyson Fick • Phone: tel +1.610.832.9710 | Upcoming Meeting: Dec. 7-11, 2014 • San Diego, Calif.