I have a bunch of samples ~500 from a continuous distribution. The objective is to generate new random samples from this distribution. How to approach this problem? Can someone give me a pointer. I am using Python so I got numpy etc.

  • $\begingroup$ You can look for the term bootstrap and see how/if it is implemented in your favorite software. $\endgroup$ – Maarten Buis Apr 10 '15 at 7:12
  • 2
    $\begingroup$ Should the generated numbers be different from the original ones? $\endgroup$ – user603 Apr 10 '15 at 7:16
  • $\begingroup$ yes, number can be anything in that range. Thanks for the link to the other post :) $\endgroup$ – Shimano Apr 10 '15 at 8:11
  • 1
    $\begingroup$ Shimano -- does that one answer your question, or are you looking for something different? $\endgroup$ – Glen_b -Reinstate Monica Apr 10 '15 at 8:18
  • 1
    $\begingroup$ As stated, i.e., only provided with 500 real numbers, there is no way you can generate new samples from that distribution. Only approximations are available. $\endgroup$ – Xi'an Apr 10 '15 at 8:41