On Sunday, November 17, 2013 4:29:29 PM UTC-5, Rich Ulrich wrote: > I would say that the population can be described by (or as) a > Poisson process, and I think other biostatisticians would, too. > Maybe some specialties use the language you suggest, but it seems > more proper, to me, to say something like "constant hazard", etc. > > Still. "constant failure rate" implies NO PEAK. That is the > situation at the asymptote, in the picture you linked. It is not > the situation that creates a peak.
I think you hit the nail on the head, Rich. The situation to which the diagram refers is a population of parts, each governed by a Poisson process, but collectively, it isn't Poisson until the oscillations die down. It is the transient response of the entire population that I was interested in.
>> Actually, it also implies exponentially decaying PDF for >> time-to-failure. Again, sorry for not being more specific. > > That would be, "exponentially decaying time-to-NEXT-failure" for the > population. Uniform distribution. Poisson. These are intimately > related.
Indeed. But at the level of individual parts. Also at the population level, but only when the oscillations die down.