|
Activity Number:
|
363
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 6, 2008 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Biometrics Section
|
| Abstract - #301826 |
|
Title:
|
An Approach to the Analysis of Spatially Correlated Multilevel Functional Data
|
|
Author(s):
|
Ana-Maria Staicu*+ and Ciprian M. Crainiceanu and Raymond Carroll
|
|
Companies:
|
University of Bristol and Johns Hopkins University and Texas A&M University
|
|
Address:
|
Department of Statistics, Bristol, International, BS8 1TW, United Kingdom
|
|
Keywords:
|
multilevel functional data ; spatial correlation ; principal compoment analysis ; kernel estimation
|
|
Abstract:
|
We describe the framework and inferential tools for hierarchical functional data where the functions at the lowest hierarchy level are spatially correlated. We present a new approach in which the hierarchical functions are modeled nonparametrically using multilevel eigenfunction bases, which appear in a multilevel functional principal component scenario, plus a weakly stationary process to account for the spatial correlation. For each level, the eigenfunction basis is estimated from the data and the functional principal component score estimates are obtained by a conditional step; a method which is conceptually simple and straightforward to implement. A second novelty of our methodology is in using kernel smoothing estimation to estimate the spatial covariance function. The proposed procedure is illustrated with a simulation study and p27 measurements data in a carcinogenesis study.
|
