Four approaches to linear robust regression analysis are presented. In the presence of outliers or bad data points in marketing data, these procedures provide formal methods to identify outliers and to reduce their influence on the final estimates of the regression coefficients. Use of these procedures in regression models is considered in two typical marketing applications and superiority of these procedures, as compared to the traditional ordinary least squares procedure, in reducing the effect of influential observations is documented. The paper concludes with suggested guidelines for their use in marketing applications.
marketing data, outliers
The files in this collection are protected by copyright law. No commercial reproduction or distribution of these files is permitted without the written permission of Southern Methodist University, Cox Business School. These files may be freely used for educational purposes, provided they are not altered in any way, and Southern Methodist University is cited. For more information, contact firstname.lastname@example.org.