|
Activity Number:
|
337
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, July 31, 2007 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #309673 |
|
Title:
|
How Many Images Are Enough? Bayesian and Non-Bayesian Approaches to Sample Size Determination
|
|
Author(s):
|
Junheng Ma*+ and Jiayang Sun and Joe Sedransk
|
|
Companies:
|
Case Western Reserve University and Case Western Reserve University and Case Western Reserve University
|
|
Address:
|
Apt D 106, Pittsburgh, PA, 15232,
|
|
Keywords:
|
posterior predictive p-value ; power calculation ; sample size ; images ; Multinomial distribution
|
|
Abstract:
|
In developing new imaging diagnostic tools, it is often of interest to compare the quality of the resulting images from the competing tools. Each image is assessed by one or several evaluators who score the images from 1 to 5. An important scientific question is how many images are needed to detect differences in the quality of these images when using these diagnostic tools. Specifically, one is interested in finding the sample sizes needed to detect a difference between the parameters from two multinomial populations. In this paper we develop three Bayesian approaches based on the ideas of posterior predictive p-values and two non-Bayesian methods using a chi-square test and a simultaneous confidence interval. Evaluations and comparisons are made using asymptotics, simulated data and a real data application.
|
