|
Activity Number:
|
313
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 4, 2009 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Physical and Engineering Sciences
|
| Abstract - #303270 |
|
Title:
|
Interior Analysis in Multiple Linear Regression
|
|
Author(s):
|
John F. Wellington*+ and Stephen A. Lewis
|
|
Companies:
|
Indiana University Purdue University Fort Wayne and Mongrel Works, Inc.
|
|
Address:
|
Doermer School of Business and Management Sciences, Fort Wayne, IN, 46805,
|
|
Keywords:
|
Multiple linear regression ; minimum sum of absolute errors regression ; post-fitting analysis ; least squares ; interior analysis ; prediction
|
|
Abstract:
|
In many applications of the single equation linear regression model, the minimum sum of absolute errors (MSAE) regression is the estimation criterion of choice. In recent time, new diagnostics and post-fit analyses have been presented. We offer a post-fit analysis that quantifies the sensitivity of the MSAE parameter estimates to variation in the data used for model estimation. We introduce software known as the Slider that facilitates the analysis. We illustrate the analysis with an example and demonstrate the functioning of the Slider that contrasts variation in the MSAE and least squares parameter estimates. We refer to the investigation as 'interior analysis' of the MSAE fit.
|
