D. Salmerón, M. D. Chirlaque López, J. A. Cano Sánchez
Integral priors for Bayesian model selection were introduced in Cano et al. (2008) and they have been further developed and applied in several papers like Cano et al. (2007a, 2007b, 2018), Cano and Salmerón (2013, 2016) and Salmerón et al. (2015). These applications range from basic parametric problems like normal and exponential models to more complex ones like binomial regression, one way random effects and ANOVA models. Here the development of integral priors for multiple comparison is presented and applied to variable selection problems in the context of linear models. Also we show how this methodology can be easily applied to variable selection for generalized linear models and nonlinear models. When comparing two models integral priors for each one are obtained by simulation, jumping between the two models and their posteriors, now this idea is extended by allowing random jumps between all the models and their corresponding posteriors.
Palabras clave / Keywords: Bayesian model selection, multiple comparison, variable selection, Bayes factors
Programado
Sesión M01 Métodos Bayesianos
30 de mayo de 2018 10:50
Sala 1
