|
Activity Number:
|
443
- SPEED: Statistical Methods and Applications in Medical Research, Risk Analysis, and Marketing Part 2
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 10, 2022 : 10:30 AM to 11:15 AM
|
|
Sponsor:
|
Section on Medical Devices and Diagnostics
|
|
Abstract #323840
|
|
|
Title:
|
Homogeneity Test for Ordinal ROC Regression and Application to Facial Recognition
|
|
Author(s):
|
Ty Nguyen* and Larry Tang
|
|
Companies:
|
University of Central Florida and University of Central Florida
|
|
Keywords:
|
Facial Regconition;
error rates;
covariate effect;
ROC curves;
AUC;
Homogeniety test
|
|
Abstract:
|
In facial recognition, ordinal scores are given by facial examiners to show confidence about whether persons in two images are the same or two different ones. Those scores can be analyzed by ROC curve to evaluate accuracy. In this talk, we propose a homogeneity test to compare performance of facial examiners. Asymptotic properties of estimated ROC curves and their corresponding AUCs within ordinal regression framework are derived as well. Moreover, behavior of difference in ROC cures (and AUCs) among examiners are investigated in detail. Confidence intervals of difference in AUCs and confidence bands of difference in ROC curves are built up to support for performance comparison purpose. Simulations are conducted on data where scores are assumed to come from binormal distribution and both categorical and continuous covariates are involved. Finally, we apply our procedure to facial recognition data to compare accuracy performance among image examiners.
|
Authors who are presenting talks have a * after their name.
