Re: Hypergeometric Question

["Rip Stauffer" <rstauffer@mn.rr.com>]

> Strictly speaking, the best model for this experiment is hypergeometric 
> distribution because the sampling plan does not include replacement. The 
> fact is, you are only taking 1/75 of the population in each run, so the 
> difference between the two models (hypergeometric and binomial) is small. 
> You will need a great many repetitions of the experiment to be able to 
> detect this difference. So I bet that you can't tell the difference in 
> practice.
I don't know how others do it, but I always have the participants dump the 
beads back into the bowl after each draw of the paddle. Doesn't that 
constitute replacement?

Best regards,

Rip Stauffer