FRIDAY, February 22, 2008
Time: 1:30 - 2:30 PM
Constant 1048

Title: Revealing Sparse Signals in Functional Data

Andrada Ivanescu
Department of Statistics
Florida State University

I will present a novel statistical method to reveal the signal in data consisting of a sample of curves, where each curve is regarded as a realization of a stochastic process. In this framework, existing methods for inference include smoothing splines and kernel estimates. Our method successfully adapts to the sparsity of the signal. The methodology involves thresholding, and the threshold level depends on the sources of variability that exist in this type of data. I will also present an application of our method to estimate the mean volatility of stock quotes in intra-day trading.