Deriving recursion formula for the moments about the mean of the Poisson distribution

336 Views Asked by At

I'm having trouble seeing how this recursion formula holds.


enter image description here


Let $a_x(\lambda) = (x - \lambda)^r \frac{\lambda^x e^{-\lambda}}{x!}$

Firstly, we have

$$\log a_x(\lambda) = r \log(x - \lambda) + x \log \lambda - \lambda - \log x!,$$

which we differentiate to get,

$$\frac{1}{a_x(\lambda)} \frac{da_x}{d\lambda} = \frac{r}{\lambda - x} + \frac{x}{\lambda} - 1$$

so the derivative is $$\frac{da_x}{d\lambda} = a_x(\lambda) \left( \frac{r}{\lambda - x} + \frac{x}{\lambda} - 1 \right) $$

Thus, the equality being proposed in the question, fully written out, is:

$$\sum_{x=0}^{\infty} (x - \lambda)^{r+1} \frac{\lambda^x e^{-\lambda}}{x!} = \lambda \left[ r \sum_{x=0}^{\infty} (x - \lambda)^{r-1} \frac{\lambda^x e^{-\lambda}}{x!} + \sum_{x=0}^{\infty} (x - \lambda)^r \frac{\lambda^x e^{-\lambda}}{x!} \left( \frac{r}{\lambda - x} + \frac{x}{\lambda} - 1 \right) \right] $$

and I do not see how we can simplify the right hand side to get the left hand side. Have I made a mistake somewhere?

1

There are 1 best solutions below

0
On BEST ANSWER

Your computations look correct so far. You just have to complete it. The RHS simplifies to

$$\sum_{x=0}^\infty \frac{\lambda^{x+1} e^{-\lambda}}{x!} \left(r (x-\lambda)^{r-1} + (x-\lambda)^r \left( \frac{r}{\lambda - x} + \frac{x}{\lambda} - 1 \right) \right) \\ = \sum_{x=0}^\infty \frac{\lambda^{x+1} e^{-\lambda}}{x!} (x-\lambda)^r \frac{x-\lambda}{\lambda} \\ = \sum_{x=0}^\infty (x-\lambda)^{r+1} \frac{\lambda^x e^{-\lambda}}{x!}.$$

In the first line, we factored out the terms $\lambda^{x+1}$, $e^{-\lambda}$, and $x!$. In the second step, we observe that the first term in parentheses cancels with the product $(x-\lambda)^r \cdot r/(\lambda - x)$. In the last step, we note the product $(x-\lambda)^r (x/\lambda - 1) = (x-\lambda)^{r+1}/\lambda$.