Bayesian Effect Fusion for Categorical and Ordinal Predictors
Sprache des Vortragstitels:
Englisch
Original Tagungtitel:
29th International Workshop on Statistical Modelling
Sprache des Tagungstitel:
Englisch
Original Kurzfassung:
We propose sparse Bayesian modelling of the effects of categorical, i.e. nominal and ordinal covariates in regression type models. Sparsity is achieved by specifying spike and slab prior distributions and posterior inference relies on MCMC methods.
For illustration we analyse Austrian data from EU-SILC 2010, where we model the effects of social and demographic characteristics on annual (personal) income.