Current students
Marian Farah,
PhD candidate in Statistics and Stochastic Modeling
Kassie Fronczyk,
PhD candidate in Statistics and Stochastic Modeling
Elizabeth Pacheco,
MSc student in Statistics and Stochastic Modeling
(co-supervised with Bruno Sanso)
Ph.D. alumni
Matthew Taddy, Assistant Professor of Econometrics and
Statistics, University of Chicago, Graduate School of Business.
Ph.D. in Statistics and Stochastic Modeling,
Spring 2008, School of Engineering, UCSC.
Dissertation title: Bayesian Nonparametric Analysis of Conditional
Distributions and Inference for Poisson Point Processes.
In his PhD thesis work, Matt studied a flexible approach to
Bayesian nonparametric modeling and inference
for conditional
densities, including development of novel modeling frameworks
for fully nonparametric
quantile regression, multivariate
regression for survival data, and semiparametric Markov
switching regression.
Moreover, Matt developed a general modeling framework for
spatial Poisson processes, including methods for
regression with individual-specific covariates (marks) and
location-specific covariates, as well as modeling for
spatial point patterns that are observed over discrete time.
Writing the corresponding papers is work in progress;
current references include:
- Taddy, M., and Kottas, A. (2008).
``Bayesian Nonparametric Modeling for Markov Switching
Regression.''
UCSC-SOE Tech Report 2008-15.
- Taddy, M., and Kottas, A. (2007).
``A Nonparametric Model-based Approach to Inference for
Quantile Regression.''
AMS Tech Report 2007-21.
Matt was also involved in a collaborative project
on statistical modeling and sensitivity analysis for
radiative transfer computer models.
Related references include:
- Morris, R.D., Kottas, A., Taddy, M., Furfaro, R., and
Ganapol, B.D. ``A Statistical Framework
for the Sensitivity Analysis of Radiative Transfer Models.''
To appear in
IEEE Transactions in Geoscience and Remote Sensing.
- Morris, R.D., Kottas, A., Furfaro, R., Taddy, M., and Ganapol,
B. (2007). ``An Analysis of the
Uncertainties in Radiative Transfer
Models Used in Remote Sensed Data Product Generation.''
Proceedings of the NASA Science Technology Conference, June 2007.
Milovan Krnjajic, Senior Statistician, National Security
Engineering Division, Lawrence Livermore National Laboratory.
Ph.D. in Computer Science, Summer 2005, School of Engineering, UCSC.
(PhD dissertation work co-supervised with David Draper).
Dissertation title: Contributions to Bayesian Statistical Analysis:
Model Specification and Nonparametric Inference.
For the part of his PhD thesis that involved Bayesian nonparametrics,
Milovan worked on Bayesian
semiparametric methodology for quantile
regression, developing Dirichlet process mixture models for
the error distribution in an additive quantile regression formulation,
including dependent Dirichlet
process modeling for quantile regression error
densities that change nonparametrically with the covariates.
Milovan also studied several classes of nonparametric models
(based on Dirichlet process priors) for
count data that arise in treatment/control experiments.
Related references include:
- Kottas, A., and Krnjajic, M.
``Bayesian Semiparametric Modeling in Quantile Regression.''
To appear in Scandinavian Journal of Statistics
(earlier version available as
AMS Tech Report 2005-06).
- Krnjajic, M., Kottas, A., and Draper, D. (2008).
``Parametric and Nonparametric Bayesian
Model Specification:
A Case Study Involving Models for Count Data.''
Computational Statistics & Data Analysis, 52, 2110-2128.
More recently, we have been working on model-based nonparametric
regression approaches that
combine nonparametric prior models
for the regression function and the error distribution. Some
early results in the context of quantile regression are reported in
M.Sc. alumni
Joel Mefford, M.Sc. in Computer Science, Fall 2005,
School of Engineering, UCSC.
Thesis title: Bayesian Nonparametric Mixtures of Weibull
Distributions With Applications to Survival Analysis.
thanos@ams.ucsc.edu
Last updated July 21, 2008