For eliminating nuisance parameters, recent literature indicates that non-Bayesian methods based on pseudo-likelihoods can be usefully incorporated into classical Bayesian analyses (see Ventura et al., 2009, and references therein) . This approach has the remarkable advantages of avoiding elicitation on the nuisance parameters and the computation of multidimensional integrals. In Bayesian model selection, when hypotheses involve unknown parameters, also Bayes Factors have the drawbacks of calling for priors on the nuisance parameters and integration. In this setting, it may be useful to resort to a pseudo-likelihood for the parameter of interest only in order to derive non genuine BF, called Pseudo-Bayes Factors (PBF). Here, it is of interest to study inferential properties of PBF throught illustrative examples, and to give a criterion to compare PBF with genuine BF, based on the frequentist risk function.
Pseudo-Bayes factors
VENTURA, LAURA
2010
Abstract
For eliminating nuisance parameters, recent literature indicates that non-Bayesian methods based on pseudo-likelihoods can be usefully incorporated into classical Bayesian analyses (see Ventura et al., 2009, and references therein) . This approach has the remarkable advantages of avoiding elicitation on the nuisance parameters and the computation of multidimensional integrals. In Bayesian model selection, when hypotheses involve unknown parameters, also Bayes Factors have the drawbacks of calling for priors on the nuisance parameters and integration. In this setting, it may be useful to resort to a pseudo-likelihood for the parameter of interest only in order to derive non genuine BF, called Pseudo-Bayes Factors (PBF). Here, it is of interest to study inferential properties of PBF throught illustrative examples, and to give a criterion to compare PBF with genuine BF, based on the frequentist risk function.Pubblicazioni consigliate
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
