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.

Let $X$ be a Poisson random variable with mean $\lambda$. How do I show that $P[X \geq \lambda] = 1/2$? Also, I was wondering what distributions have this property that the density is concentrated equally around the mean. I see that it's true for Gaussian and Uniform distributions.

share|improve this question
2  
But it's not true. If $\lambda = 1$, then $P[X \ge \lambda] = 1 - P[X=0] \neq 1/2$. –  Rahul Oct 24 '10 at 6:01
    
Not sure if $P(X \geq \lambda) = 1/2$. Take $\lambda = 1$. Then $P(X \geq 1) = 1-e^{-1}$. –  Dinesh Oct 24 '10 at 6:04

1 Answer 1

up vote 2 down vote accepted

P[X ≥ μ] = 1/2 if and only if the mean (μ) and median "coincide". This is true for Gaussian and uniform distribution, but not always true for Poisson.

There is exactly one μ between every unit interval [n, n+1) such that P[X ≥ μ] = 1/2. Some of them are: ln 2, 1.6784…, 2.674…, 3.672…, ...

share|improve this answer
    
Actually this fact was used in bounding the max load when N = Poi(n) balls are thrown into n bins. The proof goes like this: –  Naga Oct 25 '10 at 13:13
    
Let $M$ denote the max load in any bin when $n$ balls are thrown at random (and independently) into $n$ bins. Let $M'$ denote the max load in a bin when $N = Poi(n)$ balls thrown into $n$ bins. The proof goes like this: $P(M > \epsilon) <= P(M' > \epsilon | N >= n) = P(M' > \epsilon, N >= n)/P(N >= n) <= P(M' > \epsilon) / P(N>=n$. Now since $P(N >= n) = 1/2$ (as $n = E[N]$), we get the final bound as $2P(M'>\epsilon)$. –  Naga Oct 25 '10 at 13:20
    
@Naga: Obviously, that proof is not correct for small $n$. The larger $n$ is, though, the closer $P(N>n)$ gets to 1/2. Are you maybe only interested in asymptotics? –  Jens Oct 26 '10 at 15:18

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.