Student Seminars

Somak Dutta
PhD student, Department of Statistics, University of Chicago

Statistical inference and matrix-free computations for sparse spatial mixed linear models

Gaussian Markov Random fields have a charmed place in spatial statistics literature because they are adaptable to fast and uncomplicated statistical computations and facilitate complex modeling through local specifications. Applications of these models include agriculture, astronomy, image processing, climate studies, geology, epidemiology and other areas of environmental science. In this talk, I will primarily focus on spatial mixed linear models on regular arrays that provide a new perspective to nearest neighbor adjustments. I will present a novel h-likelihood method for estimating the models and introduce matrix-free computation methods that allow fast statistical computations. Furthermore, diminishing array sizes to zero, I will show how these computations provide approximate inference for the limiting continuum spatial models based on the de Wijs process, and certain robustness of the inference to changes of scale. This is a joint work with my PhD thesis advisor Dr. Debashis Mondal.

Student Seminar Archive

For questions regarding the Statistics Student Seminar or if you are interested in presenting, please contact Joonha Park(joonhap@umich.edu) or Jingshen Wang(jshwang@umich.edu).