|
Activity Number:
|
222
|
|
Type:
|
Invited
|
|
Date/Time:
|
Tuesday, July 31, 2007 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Quality and Productivity
|
| Abstract - #307780 |
|
Title:
|
A Posterior Predictive Approach to Multiple Response Surface Optimization
|
|
Author(s):
|
John Peterson and Enrique del Castillo*+
|
|
Companies:
|
GlaxoSmithKline and The Pennsylvania State University
|
|
Address:
|
, University Park, PA, 16802,
|
|
Keywords:
|
batch effects ; Bayesian Model Averaging ; Markov Chain Monte Carlo ; missing values ; noise variables ; robust parameter design
|
|
Abstract:
|
This presentation provides an overview of an approach to multiple response surface optimization that provides optimal operating conditions and also measures the reliability of an acceptable quality result. The traditional optimization approaches of overlapping surfaces or the desirability function do not take into account the covariance structure of the data nor the model parameter uncertainty. The proposed posterior predictive approach can be used with most of the current multiresponse optimization procedures to assess the reliability of a future response. This posterior predictive approach is easy to interpret and it takes into account the correlation structure of the data, the variability of the process distribution, and the model parameter uncertainty. This approach can also be extended to accommodate noise variables, batch effects, missing values, or model form uncertainty.
|
