|
Activity Number:
|
551
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Thursday, August 10, 2006 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Statistical Computing
|
| Abstract - #307602 |
|
Title:
|
Tuned and Guided Adaptive Regression by Mixing
|
|
Author(s):
|
Panayotis Giannakouros*+ and Lihua Chen
|
|
Companies:
|
University of Missouri-Kansas City and The University of Toledo
|
|
Address:
|
2051 Brookdale Road, Toledo, OH, 43606,
|
|
Keywords:
|
model combining ; adaptive regression by mixing ; model averaging
|
|
Abstract:
|
The development of Adaptive Regression by Mixing (ARM) has provided a theoretical justification for prediction-based model combining methods and demonstrated they can have superior performance under many statistical settings. However, ARM and its implementations in various statistical settings leave potential for improvement. We systematically explore the properties of prediction-based model combining, pursuing development of a superior tuned and guided prediction-based combining algorithm. We use simulations and visualization to explore and optimize key steps of the algorithm and assess the performance of the tuned and guided algorithm relative to ARM in several settings.
|
