CMStatistics 2016: Start Registration
View Submission - CMStatistics
B0381
Title: A spatio-temporal Markov switching model for the detection of influenza outbreaks Authors:  David Conesa - Universitat de Valencia (Spain) [presenting]
Ruben Amoros - Universitat de Valencia (Spain)
Antonio Lopez-Quilez - University of Valencia (Spain)
Miguel Martinez-Beneito - Centro Superior de Investigacion en Salud Publica (Spain)
Abstract: Influenza dispersion is related with climate variables and spreads person to person, which suggests a spatio-temporal evolution of the incidence. We present a spatio-temporal extension of the Bayesian Markov switching model over the differentiated rates for the detection of influenza epidemic outbreaks previously suggested. The variable of the Markov switching model represents the epidemic and non-epidemic states. The non-epidemic state differentiated rates only share a common mean for each time. The rates on the epidemic state are spatially and temporally related through Gaussian Markov random fields. This new proposal has been compared with the previous one and offers better scores in terms of the continuous rank probability score by approximate cross-validatory predictive assessment.