This is the program for the 2010 Joint Statistical Meetings in Vancouver, British Columbia.

Abstract Details

Activity Number: 525
Type: Contributed
Date/Time: Wednesday, August 4, 2010 : 10:30 AM to 12:20 PM
Sponsor: Section on Statistics and the Environment
Abstract - #308907
Title: Online Forecasting and Prediction of Spacio-Temporal Processes with Dynamic Covariance Estimation
Author(s): Dave Zes*+
Companies: University of California, Los Angeles
Address: 3134 Kallin Ave, Long Beach, CA, 90808, United States
Keywords: geostatistics ; covariogram ; kriging ; state-space ; online estimation ; statistical learning
Abstract:

In the present work we advocate a facile state-space formulation for spacio-temporal data. We assume an underlying deterministic linear process, whose coefficients are low variance states. Solving the system is reminiscent of statistical learning, where a model is "hyperparameterized" through training. Online estimation as well as retrospective analysis are much more computationally efficient than, e.g., hierarchical models, which require sequential re-fitting for state estimation -- an attractive advantage in light of the increasing availability of large geo-statistical data sets.

Method comparisons from climatological and ecological data suggest excellent performance. Extensions include non-parametric one- and two-surface covariograms, forecasting to known locations, smoothing, prediction to arbitrary locations, and moving sites. We also include video resources for illustration.


The address information is for the authors that have a + after their name.
Authors who are presenting talks have a * after their name.

Back to the full JSM 2010 program




2010 JSM Online Program Home

For information, contact jsm@amstat.org or phone (888) 231-3473.

If you have questions about the Continuing Education program, please contact the Education Department.