The multivariate beta process and an extension of the Polya tree model

摘要: We introduce a novel stochastic process that we term the multivariate beta process. The is defined for modelling-dependent random probabilities and has marginal distributions. use this to define probability model family of unknown distributions indexed by covariates. each distribution Polya tree prior. An important feature proposed prior easy centring nonparametric around any parametric regression model. implement inference survival can be adopted extend support models.