B1642
Title: Generalized spatial conditional overdispersion models: Applications and spatio-temporal extensions
Authors: Mabel Morales Otero - University of the Basque Country, Facultad de Economia y Empresa (Spain) [presenting]
Vicente Nunez-Anton - Universidad del Pais Vasco UPV/EHU (Spain)
Abstract: The purpose is to introduce and review the generalized spatial conditional overdispersion models, where regression structures are specified both for the conditional mean and for the overdispersion parameter structures. In these models, the possible spatial correlation in the data is modelled by incorporating a spatial term in these regression structures by means of a parameter that directly estimates the intensity of the spatial association. We illustrate their performance for Poisson and binomially distributed responses by fitting them to the infant mortality rates and the mothers' postnatal screening period in Colombia datasets. Moreover, we also include their comparison with the widely used BYM and BYM2 models. In addition, we propose a direct spatio-temporal extension of the spatial conditional models, where we include the spatial lag of the response variable for each time unit in the linear predictor. Finally, we also propose the temporally varying spatial lag coefficient models, which allow the coefficient for the spatial term to vary with time. In order to illustrate their behavior, we apply our proposals, for Poisson distributed responses, to the respiratory hospital admissions in Glasgow data and, for binomially distributed responses, to the Georgia low birth weight data. In each case, we compare the performance of such models with the widely used Knorr-Helds models.
