
 September 22 Lecture (Site not responding. Last check: 20071022) 
  Because, as we've seen, for reasonable size samples, the empirical cdf is a good approximation to the population cdf, using the empirical cdf to generate samples is essentially equivalent to generating samples from the original population. 
  As a result, the bootstrap distribution of a statistic should be a good approximation of the sampling distribution of that statistic. 
  In short, the bootstrap distribution of a statistic should be a reasonable approximation of the sampling distribution of that statistic, if the sample sizes are reasonably large. 
 www.unc.edu /courses/2003fall/biol/145/001/docs/lectures/Sep22.html (1076 words) 
