|
Activity Number:
|
482
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Thursday, August 7, 2008 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #302534 |
|
Title:
|
Effect of Hyperparameters in the Normal Conjugate Model on Posterior Estimates
|
|
Author(s):
|
Susan Alber*+ and J. Jack Lee+
|
|
Companies:
|
The University of Texas M.D. Anderson Cancer Center and The University of Texas M.D. Anderson Cancer Center
|
|
Address:
|
Dept. of Biostatistics Unit 447, Houston, TX, 77030, 1515 Holcombe Blvd. Unit 447, Houston, TX, 77030,
|
|
Keywords:
|
prior specification ; Bayesian ; sensitivity analysis
|
|
Abstract:
|
In Bayesian analysis prior specification is critical. We investigate the frequentist properties of point estimates in a normal conjugate model for multiple treatment groups with correlation between treatment means. Assuming that the true distribution of the data is normal with independent observations, given fixed expected value for each treatment, and a common variance across treatments, we demonstrate how the mean square error (MSE) of point estimates for treatment means and variances depend on the choice of the hyperparameters. We use our results to provide guidelines for choosing hyperparameters in data analysis, where the true data generating model is not known. We conclude that for the normal model it is feasible to choose conservative hyperparameter values that reduce the MSE of parameter estimates as compared to the standard frequentist estimates.
|
