×

Implementation of Bayesian non-parametric inference based on beta processes. (English) Zbl 0888.62034

Summary: N. L. Hjort [Ann. Stat. 18, No. 3, 1259-1294 (1990; Zbl 0711.62033)] constructs prior distributions for cumulative hazards using stochastic processes with non-negative independent increments. A particular class of processes termed beta processes is introduced there for this purpose. It is also shown that the posterior cumulative hazard is again a beta process given exact and right censored data. We develop an algorithm that enables approximate random variate generation from the posterior process using the Lévy formula for its moment generating function. Computation is illustrated using failure time data.

MSC:

62G07 Density estimation
62M05 Markov processes: estimation; hidden Markov models

Citations:

Zbl 0711.62033
PDF BibTeX XML Cite