CFE 2015: Start Registration
View Submission - CFE
A0506
Topic: Title: Bayesian variable selection in spatial autoregressive models Authors:  Jesus Crespo Cuaresma - Vienna University of Economics and Business (Austria)
Philipp Piribauer - Vienna University of Economics and Business (Austria) [presenting]
Abstract: Recently, Bayesian model averaging approaches have gained momentum in order to deal with the problem of model uncertainty by producing parameter inference unconditional on model specification issues. For spatial autoregressive models, however, standard Bayesian model averaging techniques involve the computation of Bayesian marginal likelihoods, which do not have closed form solutions. When the degree of uncertainty is large, model averaging thus leads to a severe computational burden, since the calculation of Bayesian marginal likelihoods require numerical integration techniques. Two alternative Bayesian variable selection approaches for spatial autoregressive models are put forward. Both approaches allow us to deal with the problem of model uncertainty in spatial autoregressive models in a very flexible and computationally efficient way, since the approaches can be implemented in a Gibbs sampling algorithm in a straightforward way. In a simulation study it is shown that the variable selection approaches outperform existing Bayesian model averaging techniques both in terms of computational efficiency as well as in terms of in-sample predictive performance.