WW15
Diagnosing and Treating Multicollinearity
Tue, PM
Andrew Karp
Sierra Information Services, Inc.
Email: SFBAY0001@aol.com
Abstract:
This is a half-day workshop for statisticians, data analysts and others who create and implement the results of statistical models where two or more of the continuous-level independent (predictor) variables are highly correlated with each other. This condi
tion, often called “multicollinearity” or “multivariate ill-conditioned data,” plagues the creation and interpretation of statistical models with highly correlated independent variables regardless of the characterization of the dependent (or response) vari
able as being either categorical or continuous.
In this session you will see how apply diagnostic tools available in PROCs REG, FACTOR, VARCLUS and PRINCOMP to test for the presence of collinear independent variables and potential “treatments” including
principal components regression using a combination of PROCs PRINCOMP and REG.
The materials presented in this session assume that you have a working knowledge of SAS System programming concepts and are also familiar with statistical concepts such as i
Biography:
Andrew Karp is President of Sierra Information Services, Inc. a SAS Institute Alliance Partner™ located in the historic Calfiornia wine country city of Sonoma, north of San Francisco. A 20-year user of SAS System software, Andrew is a SAS Certified Profess
ional for Version 8 of the SAS System and has presented numerous papers at SAS user group meetings in six countries. His consulting practice focuses on using the SAS System for data mining, predictive modeling and other analytic processes, as well as on de
veloping and presenting training classes on numerous SAS Software topics. He is currently writing a book to accompany the recently released SAS Learning Edition for SAS Institute’s Books by Users Press. He holds undergraduate and graduate degrees from The
George Washington University in Washington, DC.