"Everything should be made as simple as possible, but not simpler." - Albert Einstein

Seminar Announcement

Hierarchical spatial modeling of additive and dominance genetic variance for large spatial trial datasets

Dr. Sudipto Banerjee, PhD, Division of Biostatistics, School of Public Health, University of Minnesota.

Monday, February 23, 2009

4:00 pm 223 Weber


With accessibility to geo-coded locations where scientific data are collected hrough Geographical Information Systems (GIS), investigators in genetic field trials are increasingly turning to spatial process models for modelling associations and relationships over space. Over the last decade hierarchical models implemented through Markov Chain Monte Carlo (MCMC) methods have become especially popular for genetic models, given their flexibility and power to estimate models (and hence address scientific hypothesis) that would be infeasible otherwise. However, introducing spatial correlations often entails expensive matrix
decompositions whose computational complexity increases exponentially
with the number of spatial locations, rendering them infeasible for large trials.

In this talk we primarily focus upon the use of a predictive process derived from the original spatial process that projects process realizations to a lower-dimensional subspace thereby reducing the computational burden. This approach can be looked upon as a process-based approach to reduced-rank methods for "kriging". We discuss attractive theoretical properties of this predictive process as well as its greater modeling flexibility compared to existing methods. In particular, we show how the predictive process seamlessly adapts to settings with non-stationary and multivariate processes. We also discuss some pitfalls of this and other reduced-rank methods and offer remedies. A computationally feasible template that encompasses these diverse settings will be presented and illustrated. (Joint work with Andrew O. Finley, Patrick Waldmann, and Tore Ericsson.)