Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

I am not very well versed in statistics so any clarification would be appreciated. I understand the mathematical derivation of Poisson from Binomial.

I can see just from plotting various Binomial distributions where I keep p constant and increase n, that Binomial will eventually converge to a Poisson.

That fits with the derivation where n approaches infinity, but I don't see why p has to be very small? What is the intuition behind p approaching 0?

share|improve this question

1 Answer 1

The idea is that the poisson distribution is a one-parameter distribution (often $\lambda$), and that $\lambda$ corresponds to $np$ in the binomial distribution. So the limit you want is for $n$ to get larger while $np$ stays fixed - and this is why $p$ has to get smaller. Both $\lambda$ and $np$ count the number of events, which is why the correspond in this way - and the limit is taken while keeping the mean number of events constant.

share|improve this answer
    
So in a binomial distribution, p is the "true" probability of the trial (or as good as you can get with a large population size) and n is the number of trials you are conducting. In a poisson, there is no "true" probability, only the rate (lambda). Is n approach infinite and p approach 0 a good approximation for the true probability? –  damiankao May 16 '13 at 10:43
    
What you need is to look at the limit as $n \to \infty$, and set $p=\cfrac {\lambda}n$. It is not just that $p \to 0$, but that this happens in a special way. This also ensures that the variance $npq \to \lambda$, as you will want if your limit is to have any chance of being a poisson distribution. –  Mark Bennet May 16 '13 at 11:23

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.