QA350 : Approximate Bayesian inference in non-homogeneous Poisson processes with application to analysis of recurrent events
Thesis > Central Library of Shahrood University > Mathematical Sciences > MSc > 2016
Authors:
Abstarct: Due to the complexity of appropriate statistical models for Bayesian analysis of recurrent event data, we need to use sampling-baxsed methods such as MCMC algorithms. However, MCMC methods applied to these models come with a wide range of problems in terms of convergence, mixing properties and computational time. An alternative method is integrated nested Laplace approximation (INLA) method, introduced by Rue et al. (2009). The INLA method is very fast and does not suffer the same problems as MCMC. Moreover, the approximations described by INLA to be extremely accurate so that, in order for any bias to be detected, the MCMC algorithm would have to run for much longer time than it is usually done in practice. In this thesis, we assume that recurrent events occur according to a non-homogeneous Poisson process and then perform approximate Bayesian inference using INLA. We illustrate our approach using both simulated and real life data.
Keywords:
#Recurrent events #integrated nested Laplace approximation #non-homogeneous Poisson process #Cox proportional hazard model #Gaussian Markov random field Link
Keeping place: Central Library of Shahrood University
Visitor:
Keeping place: Central Library of Shahrood University
Visitor: