Short Course

Jay M. Ver Hoef has been a Biometrician with the Alaska Department of Fish and Game and an adjunct professor of statistics at the University of Alaska, Fairbanks since 1991. His fields of major statistical activities are in spatial statistics, sampling, and Bayesian statistics. He has developed an applied course on spatial statistics that he has taught throughout the United States and Europe.

Course description

Spatial statistical methods allow the modeling of data using spatial information. Spatial statistics are used in diverse subject areas ranging from environmental to biomedical applications. Examples include interpolation methods such as kriging, spatial regression, and disease-mapping. This course will cover two main areas in spatial statistics: geostatistics and lattice models.

The course will begin with an introduction to applied spatial statistics. We build on familiar concepts from linear models including regression and ANOVA, and extend these models and analyses for spatial data. Significant new implementations of software in all of these areas have occurred in recent years, including SAS, R, SPlus, WinBUGS, and ARCINFO/ARCVIEW. The course will include a lab, and a CD is provided that includes a monograph on the technical background, powerpoint presentations, all data sets that are used in the course, and all computer code that is used to analyze the data and illustrate concepts.

Course Topics

Geostatistics
    - Introduction
    - Exploratory Spatial Data Analysis (ESDA)
    - Inverse Distance Weighted
    - Variograms and spatial covariance
    - Estimating variograms and spatial covariance
    - Ordinary, Simple, Universal, Indicator, Kriging
    - Cokriging
    - Kriging neighborhoods
    - Crossvalidation
    - Geostatistical regression
    - Estimating spatial regression parameters
    - Spatial maximum likelihood and restricted maximum likelihood
    - Geostatistical regression compared to classical regression
    - Generalized linear models using a spatial covariance matrix
    - Estimating effects in spatially designed experiments
    - Geostatistical ANOVA compared to classical ANOVA
    - Model-based statistics vs. design-based statistics
    - Block kriging
    - Geostatistics compared to classical sampling

Lattice Models
    - Neighborhoods
    - ESDA for Lattice Models
    - Moran's I and Geary's c
    - CAR, SAR, and MA Lattice Models
    - Smoothing and mapping for lattice data
    - Regression for lattice models
    - Estimating lattice regression parameters
    - Estimating effects in spatially designed experiments
    - Lattice ANOVA models compared to Geostatistical and classical
    - Lattice models for sampling
    - Lattice models compared to block kriging and classical sampling

Graybill Conference
June 16-18, 2004
University Park Holiday Inn
Fort Collins, CO 80526
www.stat.colostate.edu/graybillconference
Conference POSTER
email: nsu at stat.colostate.edu Fax: (970)491-7895 Phone: (970)491-5269
Last Updated: Friday, April 09, 2004