Learn about professor Robert McCulloch’s research on exploiting potential monotonicity of the predictors during the next installment of Decision Systems Engineering Seminar Series, hosted by associate professor Rong Pan.
Seminar: Multidimensional Monotonicity Discovery with MBART
Friday, October 5th, 2018
Brickyard Engineering (BYENG) 210, Tempe campus [map]
For the discovery of a regression relationship between y and x (a vector of p potential predictors) the flexible nonparametric nature of BART (Bayesian Additive Regression Trees) allows for a much richer set of possibilities than restrictive parametric approaches. To exploit the potential monotonicity of the predictors, we introduce mBART, a constrained version of BART that incorporates monotonicity with a multivariate basis of monotone trees, thereby avoiding the further confines of a full parametric form. Using mBART to estimate such effects yields:
- function estimates that are smoother and more interpretable.
- better out-of-sample predictive performance.
- less post-data uncertainty.
By using mBART to simultaneously estimate both the increasing and the decreasing regions of a predictor, mBART opens up a new approach to the discovery and estimation of the decomposition of a function into its monotone components.
About the Speaker
Robert McCulloch is currently professor of statistics at the School of Mathematical and Statistical Sciences at Arizona State University. McCulloch’s research focuses on applied Bayesian methodology but recently has explored tree-based ensemble modeling, time series models for financial data and categorical data models for survey data with marketing applications.
He received his doctorate in statistics from the University of Minnesota in 1985. Before moving to ASU, McCulloch spent 27 years at the University Of Chicago Booth School Of Business and four years at the McCombs School of Business at the University of Texas.