|
Activity Number:
|
353
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Tuesday, August 4, 2009 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Nonparametric Statistics
|
| Abstract - #304713 |
|
Title:
|
Wavelet-Based Functional Mixed Models via DPM
|
|
Author(s):
|
Alejandro Villagran*+ and Sang-Han Lee and Marina Vannucci
|
|
Companies:
|
Rice University and Nathan S. Kline Institute for Psychiatric Research and Rice University
|
|
Address:
|
Department of Statistics, Houston, TX, 77251-1892,
|
|
Keywords:
|
Functional Mixed Models ; Discrete Wavelet Transform ; Dirichlet Process Mixture
|
|
Abstract:
|
Recently, various experimental designs in public health and bioinformatics require the use of functional mixed models (FMM), i.e., a functionalized extension of linear mixed models, conditional on modern technologies allowing researchers to record data sampled on a fine grid. Morris and Carrol (JRSS-B, 2006) developed wavelet-based functional mixed models that are flexible enough to accommodate a broad range of functional data. We attempt to extend these methods by fitting the functional mixed model in the wavelet domain and by relieving the assumption of normality on the random effect functions to any distributional form via a Dirichlet Process Mixture (DPM) prior. We use the Gibbs sampler algorithm to estimate the posterior distributions of the parameters. We illustrate this methodology with a simulated example.
|
