P. Barbillon, A. Forte Deltell, R. Paulo
Variable selection has been always a very important problem helping us to better understand and improve the process of interest. Much effort has been put in studying model selection-based variable selection methods in different statistical scenarios but, what happens in the world of mathematical models?
This is a fundamental question since mathematical models require a careful study of the uncertainty behind them which should include the study of potentially related covariates.
In this work we consider the bias function usually used to assess the gap between the model and realty as a tool to perform variable selection. In particular, we use a bias function based on the one proposed by Linkletter et al (2006) which parameters regulate the importance of the covariates in the bias. Then we implement a model selection process using models which differ in the prior choice for those parameters and compute the corresponding Bayes factors using an efficient bridge sampling procedure.
Keywords: Bayesian Analysis, Variable Selection, Model Uncertainty
Scheduled
GT8-2 Bayesian Inference
September 4, 2019 12:00 PM
I2L7. Georgina Blanes building