The first talk in the RSSNI Autumn Seminar Series is
Estimation of a model for spatial binary data

by Dr. Gabrielle Kelly, UCD, Dublin, Ireland.

Abstract
Spatial binary data where the correlations satisfy the Fréchet-Hoeffding bounds are modelled using a multivariate normal copula-based model. The binary variables are related to latent variables that have a Matérn spatial correlation via a multivariate probit model. Thus, both the marginal means of the binary variables and their spatial distances contribute to their correlations. The model yields a log-likelihood for the regression and spatial parameters that can be approximated using the numerical methods of Genz-Bretz or Madsen and this log likelihood is then optimised. Simulations demonstrate that the Madsen method is biased upward for both regression and scale parameters for small values of the Matérn shape parameter but not for larger values. The Genz-Bretz method performs well for the regression parameters but does not provide reasonable estimates for the scale parameter. The model is fitted to bTB infection data in cattle herds and wildlife badgers in Ireland. Minimum and maximum values of the binary correlations are estimated and associated Euclidean distances recorded. Our findings suggest that cattle herds infect neighbouring cattle while badgers roam over a larger area and thus have a greater infectivity range.

All welcome

 

Dr Gabrielle Kelly, UCD, Dublin, Ireland

Prof. Gilbert MacKenzie for RSS Northern Ireland Local Group