|Date/Time:||Monday, 28 Mar 2016 from 4:10 pm to 5:00 pm|
|Actions:||Download iCal/vCal | Email Reminder|
The classical Glivenko-Cantelli theorem says that for iid rv the empirical cdf converges to the true cdf uniformly on R w.p.1. In this talk we extend this to delayed regenerative sequences, delayed stationary sequences and delayed exchangeable sequences. Then we apply this to irreducible Markov chains admitting a stationary probability distribution with either a countable state space or Harris recurrent Markov chains.
This is of relevance to MCMC methodology. A key tool used is a generalisation of Polya's theorem on convergence of cdfs.
(joint work with Vivek Roy of ISU Statistics)