Abstract #301053

This is the preliminary program for the 2003 Joint Statistical Meetings in San Francisco, California. Currently included in this program is the "technical" program, schedule of invited, topic contributed, regular contributed and poster sessions; Continuing Education courses (August 2-5, 2003); and Committee and Business Meetings. This on-line program will be updated frequently to reflect the most current revisions.

To View the Program:
You may choose to view all activities of the program or just parts of it at any one time. All activities are arranged by date and time.

The views expressed here are those of the individual authors
and not necessarily those of the ASA or its board, officers, or staff.


Back to main JSM 2003 Program page



JSM 2003 Abstract #301053
Activity Number: 63
Type: Contributed
Date/Time: Sunday, August 3, 2003 : 4:00 PM to 5:50 PM
Sponsor: Section on Statistical Computing
Abstract - #301053
Title: Comparison of Prediction Accuracy for Different Predictors
Author(s): Wei Zhao*+ and Thomas J. Santner and William I. Notz
Companies: The Ohio University and The Ohio State University and Ohio State University
Address: 1958 Neil Ave., Columbus, OH, 43210,
Keywords: prediction accuracy ; empirical BLUP ; mean squared prediction error ; training data design ; computer experiments
Abstract:

We performed a simulation comparison of the predictive accuracy of cubic regression and six Empirical BLUPs (EBLUPs). The EBLUPs were obtained by estimating the correlation parameters of two correlation functions (Matern and Power Exponential) using three methods: MLE, Restricted MLE (REML), and cross-validation (XV). Each predictor was used to predict a grid of 625 points on 200 random bivariate surfaces on the unit square. Fifty surfaces were drawn using each of four mechanisms: "near-cubic'' functions, and three Krigifier surfaces with Matern correlation with shape parameters 5, 10, and 50. Each predictor was used with three different 16-point training data sets: a maximin Latin hypercube design (LHD), a Sobol design, and a D-optimal design based on the cubic model. For each combination of predictor, training data, and random mechanism, the true values were compared to the predicted values using mean squared prediction error (MSPE). We found cubic regression and the XV-EBLUPs to be worse than the other four EBLUPs and little difference between the MLE and REML EBLUPs for both of the two correlation families. Overall, the LHD has the smallest MSPE among the three designs.


  • The address information is for the authors that have a + after their name.
  • Authors who are presenting talks have a * after their name.

Back to the full JSM 2003 program

JSM 2003 For information, contact meetings@amstat.org or phone (703) 684-1221. If you have questions about the Continuing Education program, please contact the Education Department.
Revised March 2003