Terrance Dean Savitsky
Overview
Biography
Terrance Savitsky's research work focuses on Bayesian non-parametric methods to model non-linear mean response functions under GLM constructions. Savitsky's primary areas of focus in the non-linear modeling space include: 1. Gaussian process models employing covariance functions of predictors that allow the data to learn response surfaces constrained to any desired order of smoothness; 2. Dirichlet process priors under the Gaussian process formulations that capture dependences or clustering among predictors. These models implement Bayesian variable selection with a focus on high-dimensional datasets under model sparsity to improve model power / detection. Savitsky also conducts work on efficient Markov Chain Monte Carlo algorithms, including employment of adaptive methods, to promote efficient computation.
RAND® is a registered trademark. Copyright © 1994-2012 RAND Corporation.