Activity Number:
|
378
|
Type:
|
Topic Contributed
|
Date/Time:
|
Thursday, August 15, 2002 : 8:30 AM to 10:20 AM
|
Sponsor:
|
Section on Bayesian Stat. Sciences*
|
Abstract - #300873 |
Title:
|
Modeling Multiple Diffusion Curves Using Hierarchical Bayes Models
|
Author(s):
|
Tom Shively*+ and Edward George
|
Affiliation(s):
|
University of Texas, Austin and University of Pennsylvania
|
Address:
|
CBA 5.202, Austin, Texas, 78746, USA
|
Keywords:
|
Clustering ; Dirichlet process prior ; Forecasting ; High technology products
|
Abstract:
|
This paper develops a hierarchical Bayes multivariate diffusion model to model the diffusion through time of separate but related products. The parameters of the diffusion model are modeled using a Dirichlet process prior. The use of this prior allows the diffusion curves of related products to cluster and results in better estimates of each of the individual diffusion curves than would be obtained if each curve was estimated separately. The model is then used to forecast the diffusion of a new product by combining the information from the previous diffusion curves with the incoming sales data of the new product. The model is applied to diffusion curves for high technology products.
|