Seminarium i matematisk statistik
torsdagen den 30 mars 13.15 i MH:227

Antti Penttinen
Dept. of Statistics
University of Jyväskylä

Titel:
Local association tables in modelling of spatial association

Abstract:
Association between two species is considered where the
species' geographical distribution is not necessarily
homogeneous. The approach is based on log-linear modelling of
local association tables. Each component of the log-linear
presentation of the model is simultaneously modelled as mixed
models containing site specific covariates and a spatially
correlated random effect (in the sense suggested by
Besag & York & Mollié 1991). Our approach is Bayesian where the
posterior is calculated using MCMC simulation. The "local 
association tables" method is illustrated by an
environmental/ecological case study of association of two lichen
species growing on aspen trees in a primeval forest area of 25 ha.
Further possibilities and potential applications e.g. to disease
mapping and other related work will be discussed. 

The study is a joint work with M.Sc. Anne Riiali (Rolf Nevanlinna 
Institute, Helsinki) and Ph.D. Mikko Kuusinen (Ministry of the 
Environment, Helsinki).