STAT 549-001: Functional Data Analysis and Wavelets

Wavelets in Statistics and their Applications

Marina Vannucci

Spring 2008


Goals and Objectives:
This is a one-semester 3-credit course on the theory and practice of wavelets. Wavelet methods have become extremely popular in areas such as signal processing, image analysis, data compression, and statistics. This course will touch upon both theoretical and practical aspects of wavelets. Emphasis will be given to statistical modeling in the wavelet domain, particularly to Bayesian inferential procedures, and applications to real data. The course objective is to illustrate practical applications of the wavelet methods. Complex mathematical details and filtering theory of the wavelets will be only partially covered.

Instructor:
Marina Vannucci, Professor, Department of Statistics, Rice University.
Office: Duncan Hall, room 2083. Phone: 713-348-6132. E-mail: marina@rice.edu

Prerequisites:
No knowledge of wavelets is required. Some knowledge in statistics is desirable, particularly in Bayesian inference.

Textbook:
The textbook for the course is Statistical Modeling by Wavelets, by Brani Vidakovic, Wiley. See also books on wavelets on the webpage of the course.

Summary:
The course will start with a brief introduction to the wavelet theory, covering continuous and discrete wavelet transformations, connections with signal processing and Fourier analysis, constructions of wavelets and multiresolution analysis. The second part of the course will focus on wavelet-based statistical methods and applications. Topics will include smoothing of noisy signals, nonparametric function estimation and representation of stochastic processes. Emphasis will be given to Bayesian statistical modeling.
MATLAB software will be used for class demonstrations.

Tentative list of topics:
- Mathematical preliminaries and historical overview
- Continuous and discrete wavelet transformations
- Construction and properties of some families of wavelets
- Multiresolution analysis
- Overview of available wavelet software
- Wavelet shrinkage, thresholding policies, traditional and Bayesian approaches
- Wavelet regression, density and function estimation, traditional and Bayesian approaches
- Multiple curves, multivariate curve regression, hierarchical functional data
- Wavelets and time series, scalograms, variance decompositions
- Wavelet transformations of stationary processes, selfsimilarity, change point

Grades:
The course grade will be based on homeworks and on a final project. The final project can be either an analysis of new data or a critical review of the literature on aspects not fully covered in class.
Any available wavelet software can be used to carry out the final project (MATLAB, S-PLUS, MATHEMATICA, ...)

Back to the main page of STAT549: Wavelets


Copyright 2004-2008 Marina Vannucci