Short Course Sponsored by the NSF-Census Bureau Research Network

Statistics for Spatio-Temporal Data
April 17, 2013/ 8:30 a.m. - 4:00 p.m.

The course will follow the recently published book by Noel Cressie and Chris Wikle, Statistics for Spatial Data (Hoboken NJ: John Wiley and Sons, 2011). Professor Cressie will present state-of- the-art methods for spatio-temporal processes, bridging classic techniques with modern hierarchical statistical modeling concepts. From understanding environmental processes and climate trends to developing new technologies for mapping public-health data and the spread of invasive species, there is a high demand for statistical analyses of data that take spatial, temporal, and spatio-temporal information into account. The course will consider a systematic approach to key quantitative techniques for the analysis of such data, particularly hierarchical (empirical and Bayesian) statistical modeling with an emphasis on dynamical spatio-temporal models. Illustrative real-world examples will be presented throughout the course. The course is composed of four subsections:

• Overview of Statistical Modeling of Complex Data and Processes (key concepts of hierarchical modeling and overview of requisite background in spatial statistics and time series).
• Descriptive Analysis of Spatio-Temporal Data (exploratory techniques for spatio- temporal data including visualization and empirical dimension reduction).
• Fundamentals of Spatio-Temporal Modeling (spatio-temporal covariance functions, spatio-temporal kriging, relationship between process and covariance, and random-effects perspectives).
• Spatio-Temporal Dynamical Models (hierarchical perspective of data, process, and parameter models for spatio-temporal dynamical systems; parameterization approaches for dealing with the curse of dimensionality; computation).

Instructor: Dr. Noel Cressie, Distinguished Professor at the National Institute for Applied Statistics Research Australia (University of Wollongong, Australia) is the instructor. He has published extensively in the areas of statistical modeling, analysis of spatial and spatio- temporal data, and empirical-Bayesian and Bayesian methods.

Location: Census Bureau Headquarters, Suitland MD, Conference Room 4 (Suitland Metro stop). All visitors should enter through Gate 7 (Metro entrance) that morning. Breakfast and lunch is available at the Census Bureau cafeteria.

Note: Attendees should have at least a Master’s level background in probability and statistical inference, and they should have a good understanding of matrix algebra.

Registration: Contact Sandra Heineck (Sandra.L.Heineck@census.gov) to reserve a place (up to 50 students can be accommodated). Any non-citizens desiring to attend should also contact Eugene Vandrovec directly at 301.763.6418 at least a week in advance.