Sugi23.197

From sasCommunity

SUGI23.197, Poster, 3 pages, Best Contributed Paper (2nd Place)

[edit] CHEKOUT: A SAS(R) Program to Screen for Outliers

James Handsfield,

Centers for Disease Control and Prevention, Atlanta, GA.

http://www2.sas.com/proceedings/sugi23/Posters/p197.pdf

http://www.sascommunity.org/wiki/ChekOut.sas

ABSTRACT

At various times, most researchers are faced with the need to screen for outliers. Many statistical programs will do this with more or less ease. SAS has two PROCs which will calculate several outlier statistics. They can produce a report and output data set which require further manipulation to use. This data manipulation opens the door of opportunity for program errors which might be difficult to detect. The more statisticians can rely on known code, the more comfortable they can be with the outcome. CHEKOUT uses several macros to write the value detected as an outlier and its associated RSTUDENT statistic in the output data set from PROC GLM to a text file, OUTLIER.SAS. Also written to this file is the default option to convert the value to missing, and the file is printed as an outlier screening report. OUTLIER.SAS is easily reviewed with a text editor or in SAS , modified if necessary, and used with an %INCLUDE statement in a subsequent DATA step to remove or convert the outliers in an analysis program. Users of CHEKOUT should have knowledge of multivariable statistics and some experience with SAS macro language.