I'm reading a paper that uses a Poisson process to model real world events. The authors mention "finite window effects". What are finite window effects?

Here is quote from the paper where the authors first mention the term:

If the data come from a Poisson process, then a histogram of inter-event times will be roughly uniform when looking at a short inter-event time window due to minimal finite window effects. This is because, with an infinite window, inter-event times from a Poisson process are uniformly distributed.

  • 2
    $\begingroup$ It would help if we could have more context. Could you provide an excerpt from the paper? $\endgroup$ – Michael Chernick Aug 17 '12 at 20:27
  • 1
    $\begingroup$ I don't know what they are, but a bit of Googling found [this](dl.acm.org/citation.cfm?id=2239800}. Does that help? $\endgroup$ – Peter Flom Aug 17 '12 at 20:29
  • $\begingroup$ My first guess as to "finite window effects" would correspond to the article Peter Flom found or to what I'd refer to as "boundary effects". $\endgroup$ – Wayne Aug 17 '12 at 20:34
  • 1
    $\begingroup$ @MichaelChernick, sorry about that, context added! $\endgroup$ – slayton Aug 17 '12 at 20:38
  • 4
    $\begingroup$ That's some incredibly sloppy wording on the part of the authors. $\endgroup$ – cardinal Aug 17 '12 at 20:41

This is a guess but an educated one. Since the authors are referring to a histogram of interarrival times they might be referring to a smoothed version of the histogram. A kernel density estimate is one way to smooth a histogram. The bandwidth of the kernel is called the window.

Based on the article that Peter Flom linked I have a little more confidence that my guess is correct. The article deals with spectral density estimates and the rectangular and Hanning windows are particular shaped kernels.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.