Dismiss Notice
Join Physics Forums Today!
The friendliest, high quality science and math community on the planet! Everyone who loves science is here!

Normally distributed numbers

  1. Feb 16, 2012 #1
    Any ideas how to generate in any software R, Matlab etc normally distributed random numbers by one condition so they will be integers like 1, 2,3...and not decimal values.
    I tried to round up the generated decimal values but in my guess normality is doubted after that.
  2. jcsd
  3. Feb 16, 2012 #2
    You will have to be clearer than this. A distribution which only takes on integers can never be normal.

    What is it you really want??
  4. Feb 16, 2012 #3
    I mean I need to generate some sample data like 9.00 AM 9.04 AM 9.07 AM but with condition that inter-arrival times are normally distributed.
    That's why I need some numbers like 4 (9.04AM -9.00AM ) etc normally distributed.
    Any help on this?
  5. Feb 16, 2012 #4
    Arrivals are continuous. Your numbers will be real. You can round them if you like but if you do that then the approximation to normal depends on the parameters of the distribution.

    I don't know what you're trying to do, but has it occurred to you that modeling arrivals with a normal distribution allows for the possibility of an arrival occurring before the previous arrival? You might want to consider that.
  6. Feb 16, 2012 #5
    Can you please explain something more about?
  7. Feb 16, 2012 #6
    I don't know which part you're referring to. For the first, just consider a normal density with s.d.=.01 vs. one with a very large s.d. If you draw from the first and round everything to the nearest integer then you basically end up with the mean for virtually all of your points. In the second case you could round to the nearest integer and still "approximate" normal, whatever that means in your context.

    As to the second point, inter-arrival times are positive. T1 is the time from zero to the first arrival, T2 is the time from T1 of the second arrival, etc. So the times are usually described with a density that is zero for t<0 such as an exponential. If you use a normal distribution then you permit negative values and risk that, for example, T2<0, and your second arrival occurred prior to your first. You would take this into account in the context of what exactly you are trying to do.
Share this great discussion with others via Reddit, Google+, Twitter, or Facebook