Brian Reich

A hierarchical max-stable spatial model for extreme precipitation

Extreme environmental phenomena such as major precipitation events manifestly exhibit spatial dependence. Max-stable processes are a class of asymptotically-justified models that are capable of representing spatial dependence among extreme values. While these models satisfy modeling requirements, they are limited in their utility because their corresponding joint likelihoods are unknown for more than a trivial number of spatial locations, preventing, in particular, Bayesian analyses. In this paper, we propose a new random effects model to account for spatial dependence.

Thursday, November 29, 2012 - 3:30pm
Type: 

Donald Richards

Return Optimization Securities, and Other Remarkable Structured Financial Products

We analyze the mathematical foundations of three types of structured financial products: return optimization securities, yield magnet notes, and reverse exchangeable notes. These products were sold widely to retail "investors" in the mid-2000s. On the basis of their mathematical structure, we infer that these products could provide positive returns to a purchaser only if the stock market had continued on an enormous upward climb for most or all of the holding period.

Thursday, November 15, 2012 - 3:30pm
Type: 

Ling Lan

Georgia Health Sciences University

Nonparametric Regression for Current Status Multistate Models with Informative Cluster Size

We propose nonparametric estimators for the state occupation probabilities in a given state adjusting for the informative cluster size and one covariate at a time in a multistate model. This is a non-trivial problem since the state occupied is determined at a single inspection time for each subject and a group of subjects belongs to a cluster where cluster size is informative to their state status.

Thursday, November 8, 2012 - 3:30pm
Type: 

Eric Chicken

Nonparametric Change Point Detection and Estimation in Sequential Data

Many modern processes are capable of generating rich and complex data records not readily analyzed by traditional techniques. A single observation from a process might consist of n pairs of bivariate data that can be described via some functional relation (for example, a sequence of radar reflection signals measured over time). Or, each observation in a process may be a sample of data from some distribution. Methods are proposed here for detecting changes in such sequences from some known or estimated nominal state.

Thursday, November 1, 2012 - 3:30pm
Type: 

Eric Feigelson

Astrostatistics

This talk provides an introduction to challenging statistical problems arising in the study of celestial objects: planets, stars, galaxies and the Universe as a whole. We start with a review of the close historical links between astronomy and statistics, from the ancient Greeks through Laplace and Gauss. However, the communities diverged during the 20th century, developing into a poor state with great needs for advanced methodology but weak links between the fields. This is ameliorating today with a vibrant subfield of astrostatistics.

Thursday, October 25, 2012 - 3:30pm
Type: 

Jack Schuenemeyer

Modeling gas hydrate resources – a statistician’s perspective

Gas hydrate, which is essentially methane in ice, is a potentially important worldwide energy resource. There is evidence of significant in-place gas hydrate resources in offshore deepwater areas of the world. Southwest Statistical Consulting under grants from U.S Bureau of Ocean Energy Management (BOEM) is developing a mass-balance cell-based model using stochastic simulation to obtain estimates of in-place gas hydrate resources in U.S. Federal off-shore areas.

Thursday, October 11, 2012 - 3:30pm
Type: 

Chuanshu Ji

Probability Approximations Schemes for Generalized Black-Scholes Pricing Formulas

Asset pricing and volatility modeling take a center stage in financial econometrics. This talk introduces a new method that helps calibration of stochastic volatility models via Markov chain Monte Carlo (MCMC) Bayesian inference based on returns and option data. With the presence of high-dimensional latent volatility processes, numerical integration for computing option prices is required at every time point and every iteration of MCMC.

Thursday, October 4, 2012 - 3:30pm
Type: 

Andreas Artemiou

SAMSI

Utilizing machine learning for sufficient dimension reduction

Sufficient dimension reduction (SDR) ideas are used for supervised dimension reduction in regression problems. Support Vector Machine (SVM) algorithms belong to the class of machine learning techniques which are used for classification. In this talk we discuss Principal Support Vector Machine (PSVM) a method which utilizes SVM to achieve sufficient dimension reduction. PSVM has several advantages over existing methodology for sufficient dimension reduction, with the most important one being the fact that we can do linear and nonlinear dimension reduction under a unified framework.

Thursday, September 27, 2012 - 3:30pm
Type: 
306 Statistics Building

James Livsey

Clemson University

Small integer time series via discrete renewal processes

Discrete renewal processes are ubiquitous in stochastic phenomenon. In this talk constructing a discrete process where renewals are more (or less) likely during specified seasons is of specific interest. For example thunderstorms in the Southern United States can take place at any time in the year, but are most likely during the summer. Hurricanes, tornadoes, and snowstorms are other meteorological count processes obeying periodic dynamics. Rare disease occurrences, accidental deaths, and animal sightings are non-meteorological examples of count phenomenon following a periodic structure.

Thursday, September 20, 2012 - 3:30pm
Type: 
Room 306, Statistics Building

Martin Klein

US Census Bureau

Statistical Analysis of Noise Multiplied Data Using Multiple Imputation

A statistical analysis of data that have been multiplied by randomly drawn noise variables in order to protect the confidentiality of individual values has recently drawn some attention (Nayak, Sinha, and Zayatz, 2011; Sinha, Nayak, Zayatz, 2012). If the distribution generating the noise variables has low to moderate variance, then noise multiplied data have been shown to yield accurate inferences in several typical parametric models under a formal likelihood based analysis (Klein, Mathew, and Sinha, 2012).

Thursday, September 13, 2012 - 3:30pm
Type: 
306 Statistics Building

Pages

Subscribe to RSS - Colloquium