Activity Number:
|
389
|
Type:
|
Contributed
|
Date/Time:
|
Thursday, August 15, 2002 : 8:30 AM to 10:20 AM
|
Sponsor:
|
Biopharmaceutical Section*
|
Abstract - #301306 |
Title:
|
An Algorithm for the Recursive Partitioning of Multivariate Binary Responses and Its Application in a QSAR Study
|
Author(s):
|
Jiuzhou Wang*+ and Leonard Stefanski and Lei Zhu and Marc Genton
|
Affiliation(s):
|
North Carolina State University and North Carolina State University and GlaxoSmithKline, Inc. and North Carolina State University
|
Address:
|
Campus Box 8203, Raleigh, North Carolina, 27695-8203, USA
|
Keywords:
|
recursive partitioning ; QSAR ; multivariate binary data ; classification tree
|
Abstract:
|
In many applications, multivariate responses are of great interest. When a small number of covariates are associated with these responses, and the responses are continuous measurements, both parametric and nonparametric models are available in the literature. With large number of covariates, many of the parametric models may not work well, due to dimensionality and nonlinearity. Recursive partitioning methods have attractive features in handling this kind of problem. The objective of this work is to investigate the use of recursive partitioning algorithm for multivariate categorical responses. We will focus on the situation where both the response and covariates are binary. The methodology can be easily generalized to the case in which the responses are categorical and the covariate are ordered numerical. A real-world structure-activity data set is analized with the proposed algorithm. The result is compared to the continuous counterpart of the proposed algorithm.
|