A Nonparametric Test of Independence and Its Application in Digital Image Quality Assessment

Document Type

Presentation Abstract

Presentation Date

5-3-2010

Abstract

Statistical tools to detect general relationships between variables are commonly needed in various practices including image similarity assessment. Correlation, regression, and copula based methods provide popular tools for such purpose with certain limitations. In this talk, I will present a new nonparametric test of independence between the response variable, which can be discrete or continuous, and a continuous covariate after adjusting for heteroscedastic treatment effects. The method involves first augmenting each pair of the data for all treatments with a fixed number of nearest neighbors before a test statistic is constructed as the difference of two quadratic forms. The asymptotic distribution of the proposed test statistic is obtained under the null and local alternatives. Example applications on copulas will be given. Numerical studies show that the new test procedure has robust power to detect nonlinear dependency in presence of outliers that might result from highly skewed distributions. The application of the new test in digital image quality assessment shows that this test provides a better summary on image structure information loss compared to popular image similarity measures.

The content of this talk involves joint work with Siti Tolos, Suojin Wang, Diego Maldonado, and Sharad Silwal.

Additional Details

Monday, 3 May 2010
3:10 p.m. in Math 103
4:00 p.m. Refreshments in Math Lounge 109

This document is currently not available here.

Share

COinS