Department of Statistics Seminar
North Carolina State University

presents

Dr. Tilmann Gneiting

University of Washington

"Nonseparable, stationary covariance functions for space-time data"

ABSTRACT

Geostatistical approaches to spatio-temporal prediction in environmental science, meteorology, and related disciplines largely depend on suitable analytic covariance models. This talk proposes general classes of nonseparable, stationary covariance functions for space-time processes. The approach is based on classical results in Fourier analysis but avoids Fourier inversion. The model parameters allow for physical interpretation and include a space-time interaction parameter. Special emphasis is put on the modeling of dynamic processes such as preferred wind directions or ocean currents, and strategies for model fitting are illustrated using wind data from Ireland.

Friday, October, 04, 2002

3:35 - 4:35 pm

206 Cox Hall

Refreshments will be served on the second floor of Dabney Hall (left of Room 222) at 3:00 pm.