Multiscale Bayesian survival analysis

We consider Bayesian nonparametric inference in the right-censoring survival model, where modeling is made at level of hazard rate. derive posterior limiting distributions for linear functionals hazard, and then ‘many’ simultaneously appropriate multiscale spaces. As an application, we Bernstein–von Mises theorems cumulative functions, which lead to asymptotically efficient confidence bands these quantities. Further, show optimal contraction rates terms supremum norm. In medical studies, a popular approach model hazards priori as random histograms with possibly dependent heights. This more general classes arbitrarily smooth prior are considered applications our theory. A sampler provided histogram posteriors. Its finite sample properties investigated on both simulated real data experiments.