Re: st: Producing random numbers from a kernel density

[Marcello Pagano <pagano@hsph.harvard.edu>]

Depends how many numbers you want to generate, how much accuracy you 
want etc., but a quick and (not so) dirty method is to (1) make sure you 
have a decent density estimate (such as eqprhistogram ;-) ) that is 
always non-negative and integrates to one, then (2) calculate the 
distribution function of this density at as fine a grid as you want, 
then (3) generate a uniform random number, and (4) use the probability 
integral transform (which says that your distribution function is 
uniformly distributed) to get your variate Voila.

m.p.


Peter Willem� wrote:


> Dear all,
> I am trying to draw random numbers from an arbitrary distribution. One 
> idea that occurred  to me was to use a kernel density estimator 
> (-kdensity-), but I have no idea how to produce random numbers from 
> the result. All suggestions are welcome.
> Best,
> Peter
*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/