Activity Number:

133
 Statistical Methods for Functional Data

Type:

Contributed

Date/Time:

Monday, July 29, 2019 : 8:30 AM to 10:20 AM

Sponsor:

Section on Nonparametric Statistics

Abstract #304338

Presentation

Title:

A New Metric for Estimating Noise in Functional Data

Author(s):

Subhrangshu Nandi* and Michael Abott Newton

Companies:

Amazon and University of Wisconsin  Madison

Keywords:

functional data;
noise;
genomics

Abstract:

In functional data analysis, where both amplitude and phase variability affect individual trajectories, noise measures like pointwise variance are not appropriate because adjacent points are not independent of each other. While there exists techniques for estimating similarity between curves, there is no measure for quantifying noise in a sample of curves. We propose a novel method of quantifying the noise in a sample of noisy realizations of an underlying smooth function. The noise can be in either of the two axes. We use this measure in the analysis of graylevel intensity measurements derived from largescale experiments involving long DNA molecules where we develop an “iterated curve registration” technique to extract signal from very noisy data.
