*********************** Course Announcement ****************************

Spatial Models (STAT 597A), Spring 2010
Meeting times are Mondays and Wednesdays: 10:10am-11:25am.
Instructor: Murali Haran (mharan@stat.psu.edu)

---------------------------------------------------------------------------
Detailed description:

Spatial data appear in a large number of disciplines including
ecology, environmental statistics, public health, forestry,
geosciences and demography. A knowledge of spatial or spatio-temporal
modeling is therefore often key in these disciplines. While the focus
of this course will on spatial models, several topics covered are much
more general applicable to areas like machine learning and
classification, including Gaussian process models, Markov random
fields, point processes, Bayesian inference and Monte Carlo/Markov
chain Monte Carlo methods.

This course will cover modern approaches for modeling spatially
dependent data. Coverage will include geostatistical (point level) and
lattice (areal) data and spatial point processes. Both frequentist and
Bayesian approaches will be discussed, though the emphasis will be on
modern hierarchical Bayesian methods. The course will begin with
overviews of both spatial data analysis methods and Bayesian
methodology and computing. Some of the topics to be covered through
the remainder of the course include kriging, Gaussian random fields,
spatial generalized linear models, spatial point processes
(e.g. Poisson, Cox, Gibbs processes), and models for univariate and
multivariate spatial and spatiotemporal data. Advanced topics may
include machine learning using Gaussian processes, methods for
combining information from deterministic and stochastic models
(e.g. models from climate science, disease dynamics), nonstationary
spatial processes, new computational approaches, and other topics
suggested by students. Current research problems and methodology will
be covered whenever possible throughout the course.

A graduate level understanding of basic probability and mathematical
statistics is assumed, as is familiarity with basic computer
programming. Prior knowledge of stochastic processes and Monte Carlo
methods (as covered in STAT 515) will be very helpful. Some homework
will be assigned to help students become familiar with relevant theory
and computation but most of the course grade will be based on a
substantial semester-long course project.

Grads, post-docs and others have audited or sat in on the course in
past years; anyone interested in attending is therefore most welcome
to contact the instructor.

Course website: http://www.stat.psu.edu/~mharan/spatial/spatial.html