host: Dr Janine Illian

refID: 956

hide details

The first part of this talk is based on joint work with Victoria Landsman (Dept of Statistics) and with Yohay Carmel and Ronen Kadmon (Dept of Evolution, Systematics and Ecology) of the Hebrew University.  I will discuss a method we developed (Analyzing Spatially Distributed Binary Data Using Independent-Block Estimating Equations, Biometrics 2007) to analyze a large (6,000 pixels) set of spatially distributed data, in which a binary response was used to describe vegetation changes in a natural forest area.  I will first discuss the computational difficulties encountered when we tried to apply the Composite Likelihood approach of Heagerty and Lele (A Composite Likelihood Approach to Binary Spatial Data, JASA 1998).  I will then describe our alternative approach, which is essentially a GEE estimator in which the working covariance matrix is defined by breaking the region into disjoint blocks and modelling the spatial dependence within the blocks, while assuming independence between the blocks.  

  In the second part of the talk I will discuss some possible additional questions arising from this work: (1) comparing our approach with other methods for handling relatively large sets of spatial data; (2) extending our method from binary to ordinal responses; and (3) in the context of a Generlized Linear Model with spatial dependence defined by latent normal variables, how to graphically examine the form of the spatial autocorrelation on the latent scale.

host: Dr Janine Illian

refID: 951

hide details