Nonhomogeneous Poisson processes and logconcavity (Q2709779)

The authors consider and compare various ways logconcavity occurs in the distributions of random variables and processes. In an Appendix interrelations between the property to be logconcave for the hazard rate function, the cumulative hazard rate function, the probability density, and the survival function of a random variable are proved.NEWLINENEWLINENEWLINEThe central topic of the paper are results about logconcavity of counting processes, their epoch times, and their interepoch times. The counting processes of interest here are the non-homogeneous Poisson process, the relevation counting process (which models some minimal repair processes with non identically distributed life times), and discrete time 1-step counting processes with independent, but not identically distributed jumps. Results are of the prototype form: In a nonhomogeneous Poisson process all epoch times have logconcave densities if and only if the density and the cumulative hazard rate function of the first epoch time are logconcave. If the rate function of a nonhomogeneous Poisson process is logconcave, then all interepoch times have logconcave densities.