Generalized P-Value for Two-Sample Functional Data Comparison

• Activity Number: 544
• Type: Topic Contributed
• Date/Time: Wednesday, August 6, 2014 : 2:00 PM to 3:50 PM
• Sponsor: IMS
• Author(s): Yixuan Qiu*+ and Lingsong Zhang
• Companies: Purdue University and Purdue Univeristy
• Keywords: generalized p-value ; functional data ; mean comparison ; high-dimensional testing ; Type I error

Abstract

Hypothesis testing lies in the central part of statistical inference, while it is still challenging in functional data analysis because of the infinite dimension nature. In this talk we transform the test problem of functional means into a high-dimensional two sample comparison problem. By incorporating the concept of generalized p-value, we are able to derive a parameter-free test variable. Simulation studies show that the proposed testing approach has a Type I error not greater than the given significance level. Comparisons between our method and other existing methods are reported in terms of size and power. We also apply the proposed method to real data, indicating the effectiveness of the test variable to detect difference in functional curves.