6

I'm modeling a factory with orders arriving independently of each other. I need a series of actual time stamps of arrivals. I've been told that everyone uses the Poisson distribution to model arrivals. Fine. But what I can't understand is how getting Poisson time stamps differs from getting time stamps with a uniform distribution.

Right now I plan all the arrivals before the simulation starts by choosing uniformly distributed random numbers over the total time span. I sort this list and those are my arrival times. I measure the inter-arrival times and I find that they have an exponential distribution. But I am told that inter-arrival times of Poisson arrival times have an exponential distribution, too.

So what's up? If they have the same distribution of inter-arrival times are Poisson and uniform two names for the same distribution? Is this a discrete vs. continuous distinction?

What is the property of a set of arrival times chosen with Poisson that differs from a set chosen with uniform? Are there fewer small gaps? Fewer big gaps? Something I'm not thinking of?

vigos
  • 100
  • 1
  • 1
  • 9
user7392
  • 261
  • 2
  • 8
  • 4
    For a Poisson process, the times *between* events follow an exponential distribution. If you pick a time window over which to look at it, the count of events, $n$, follows a Poisson distribution, & given $n$ the time *of* events follows a uniform distribution. – Scortchi - Reinstate Monica Jun 29 '17 at 16:30

1 Answers1

10

@scortchi has the right answer. To summarize:

  • The arrival time stamps are uniformly distributed.
  • The inter-arrival times are exponentially distributed.
  • The count of arrivals per uniform time period is Poisson distributed.
user7392
  • 261
  • 2
  • 8