Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!

Supported by

Jitter from exponential distribution

Dear OS community :smile:

I'm trying to create a jitter randomize according to a non-symmetric exponential distribution (truncated at 4000 msc). I need the jitter to randomize between 500 msc till 4000 msc and average of 1000 msc.

Any ideas?



  • edited April 2018

    Hmm, no idea what a jitter time is! I can maybe help a bit with the distribution. The below will generate a truncated exponential distribution between 0-3500ms using the inverse cdf method, then add on 500ms. This almost gives you what you want. The mean will be a little shy of 1000 (990 something) because you are cutting a tiny bit of the tail off.

    import numpy as np
    #0.5 is 500ms, desired mean before we add on the initial 500
    p= 1 - np.exp(-1/0.5)
    #generate uniform random numbers
    r=np.random.rand( 100000)
    # N is the highest delay allowed (in seconds). 3500ms until we add on initial 500 ms at theend
     #x1000 to return in milliseconds as opensesame expects.
    gen_trunc=np.log(1-r*(1-(1-p)**N))/np.log(1-p) * 1000
    #add on 500 to get max 4000 and mean (almost) 1000
    jitter = gen_trunc + 500
    print np.mean(jitter)
    import matplotlib.pyplot as plt
  • Wow. Thank you so much!

    How can I randomly sample the function you've created?

  • Sorry for slow reply, the variable jitter is a big array of random numbers drawn from the distribution you wanted, so just take as many numbers out as you need starting from 1 to 100,000


  • I'll try that.. Thanks!!!!!

Sign In or Register to comment.