Online Program Home
  My Program

Abstract Details

Activity Number: 335 - SPEED: Reliable Statistical Learning and Data Science
Type: Contributed
Date/Time: Tuesday, August 1, 2017 : 10:30 AM to 12:20 PM
Sponsor: Biometrics Section
Abstract #323779 View Presentation
Title: A New Multivariate Measure of Agreement in Method-Comparison Studies
Author(s): Sasiprapa Hiriote* and Vernon M Chinchilli
Companies: Department of Statistics, Faculty of Science, SILPAKORN UNIVERSITY and Department of Public Health Sciences, Penn State Hershey College of Medicine
Keywords: Agreement ; Matrix-based concordance correlation coefficient ; Method-comparison studies ; Monte Carlo Simulation ; Multivariate ; U-statistics
Abstract:

In method-comparison studies, the overall level of agreement between two different methods with respect to multivariate data is often required. In this work, a new multivariate measure of agreement is constructed based on two expected matrices. This multivariate measure is proven to have all the properties needed for assessing agreement between two p-dimensional vectors. Moreover, an estimator of the multivariate measure is developed based on U-statistics. It can be proven that this estimator is asymptotically unbiased , consistent, and has an asymptotic normal distribution. The finite-sample properties of the proposed estimator are investigated compared to those of the matrix-based concordance correlation coefficient estimator using a Monte Carlo simulation. The results indicate that the proposed estimator performs much better in terms of relative bias in most scenarios. Moreover, the proposed multivariate agreement estimator has smaller relative MSE than the matrix-based concordance correlation coefficient estimator. Finally, real multivariate data from clinical and chemical studies are analyzed to demonstrate the methodology.


Authors who are presenting talks have a * after their name.

Back to the full JSM 2017 program

 
 
Copyright © American Statistical Association