Likelihood that an observation from a Poisson distribution takes an odd value

552 Views Asked by At

Show that the likelihood that an observation from a Poisson ($\lambda$) distribution takes an odd value (i.e. 1, 3, 5,...) is $\frac{1-{e^-}^{2\lambda}}{2}$.

Since likelihood is given by $L(\lambda)=\prod_{i=0}^n f(X;\lambda)$, if we write probability for each $X=2n+1$ (odd values) i.e. $P(X=1).P(X=3)$ and so on. The Result is not achievable. I mean, what is in the question is Probability or Likelihood function which we calculate to find estimators.

1

There are 1 best solutions below

0
On

This is actually an interesting computation, and it relies on the hyperbolic functions.

The pdf for a Poisson$(\lambda)$ random variable $X$ is given by:

$$P(X = k) = \frac{e^{-\lambda}\lambda^k}{k!}$$

and we want to compute the likelihood or probability that $X$ is odd. Any odd integer can be represented as $2n+1$ for some $n \in \mathbb{N}$. So we can use the pdf to write:

$$P(X = 2n+1) = \frac{e^{-\lambda}\lambda^{2n+1}}{(2n+1)!}$$

But we need to do this for every $n$ to get the desired probability so now write it as an infinite sum.

$$P(X \text{ is odd}) = \displaystyle \sum_{n = 0}^{\infty}\frac{e^{-\lambda}\lambda^{2n+1}}{(2n+1)!} =e^{-\lambda} \displaystyle \sum_{n = 0}^{\infty}\frac{\lambda^{2n+1}}{(2n+1)!} $$

Now use the Maclaurin series for the hyperbolic sine 'sinh':

$$\text{sinh } x = \displaystyle \sum_{n = 0}^{\infty} \frac{x^{2n+1}}{(2n+1)!}$$

$$P(X \text{ is odd}) = e^{-\lambda} \text{sinh }\lambda$$

Now use another clever identity for the hyperbolic sine, namely redefine it in terms of the exponential function:

$$\text{sinh }x = \frac{e^x - e^{-x}}{2}$$

Plug in and solve:

$$P(X \text{ is odd}) = e^{-\lambda} \text{sinh }\lambda = e^{-\lambda}\Big(\frac{e^\lambda - e^{-\lambda}}{2}\Big) = \frac{1 -e^{-2\lambda}}{2} $$