Need help simplifying a summation with binomials

AI Thread Summary
The discussion focuses on proving a summation involving binomials and independent Poisson random variables. Participants explore the use of generating functions and properties of exponential functions to simplify the expression. They emphasize that the moment-generating function of a Poisson distribution can aid in demonstrating that the sum of two independent Poisson random variables results in another Poisson random variable with a mean equal to the sum of the individual means. The conversation also clarifies the relationship between generating functions and the independence of random variables. Overall, the thread provides insights into the mathematical techniques necessary for the proof.
Eclair_de_XII
Messages
1,082
Reaction score
91

Homework Statement


"Prove that ##\sum_{n=0}^\infty s^n e^{-\lambda} \frac{\lambda^n}{n!}\sum_{m=0}^\infty s^m e^{-\mu}\frac{\mu^m}{m!}=\sum_{m+n=0}^\infty s^{n+m} e^{-(\lambda+\mu)} \frac{(\lambda + \mu)^{m+n}}{(m+n)}!##

Homework Equations


Binomial theorem: ##(x+y)^n=\sum_{k=0}^n x^ky^{n-k}##
Vandermonde's identity: ##\binom {n+m} m =\sum_{k=0}^m \binom n k \binom m {m-k}##

The Attempt at a Solution


##\sum_{n=0}^\infty s^n e^{-\lambda} \frac{\lambda^n}{n!}\sum_{m=0}^\infty s^m e^{-\mu}\frac{\mu^m}{m!}=e^{-(\lambda+\mu)}\sum_{n=0}^\infty \sum_{m=0}^\infty s^{n+m} \frac{\lambda^n \mu^m }{m!n!}=e^{-(\lambda+\mu)}\sum_{n=0}^\infty \sum_{m=0}^\infty s^{n+m} \binom {m+n}{n} \frac{\lambda^n \mu^m }{(m+n)!}##
##=e^{-(\lambda+\mu)}\sum_{n=0}^\infty \sum_{m=0}^\infty s^{n+m} \sum_{k=0}^n \binom m k \binom n k \frac{\lambda^n \mu^m }{(m+n)!}##

I'm afraid to go any further, because it won't get me my ##(\lambda + \mu)^{n+m}## term. If anyone has any pointers on what I should do next with this expression (or giving me another expression), or an alternate way to prove using probability-generating functions, that the sum of two independent Poisson r.v.'s have mean equal to the sum of each individual mean, then that would be much appreciated.
 
Last edited:
Physics news on Phys.org
Binomial formula?
 
Eclair_de_XII said:

Homework Statement


"Prove that ##\sum_{n=0}^\infty s^n e^{-\lambda} \frac{\lambda^n}{n!}\sum_{m=0}^\infty s^m e^{-\mu}\frac{\mu^m}{m!}=\sum_{m+n=0}^\infty s^{n+m} e^{-(\lambda+\mu)} \frac{(\lambda + \mu)^{m+n}}{(m+n)}!##

Homework Equations


Binomial theorem: ##(x+y)^n=\sum_{k=0}^n x^ky^{n-k}##
Vandermonde's identity: ##\binom {n+m} m =\sum_{k=0}^m \binom n k \binom m {m-k}##

The Attempt at a Solution


##\sum_{n=0}^\infty s^n e^{-\lambda} \frac{\lambda^n}{n!}\sum_{m=0}^\infty s^m e^{-\mu}\frac{\mu^m}{m!}=e^{-(\lambda+\mu)}\sum_{n=0}^\infty \sum_{m=0}^\infty s^{n+m} \frac{\lambda^n \mu^m }{m!n!}=e^{-(\lambda+\mu)}\sum_{n=0}^\infty \sum_{m=0}^\infty s^{n+m} \binom {m+n}{n} \frac{\lambda^n \mu^m }{(m+n)!}##
##=e^{-(\lambda+\mu)}\sum_{n=0}^\infty \sum_{m=0}^\infty s^{n+m} \sum_{k=0}^n \binom m k \binom n k \frac{\lambda^n \mu^m }{(m+n)!}##

I'm afraid to go any further, because it won't get me my ##(\lambda + \mu)^{n+m}## term. If anyone has any pointers on what I should do next with this expression (or giving me another expression), or an alternate way to prove using probability-generating functions, that the sum of two independent Poisson r.v.'s have mean equal to the sum of each individual mean, then that would be much appreciated.

Does your formula ##(x+y)^n=\sum_{k=0}^n x^ky^{n-k}## work for ##n = 2## or ##n = 3?##
 
Eclair_de_XII said:
or an alternate way to prove using probability-generating functions, that the sum of two independent Poisson r.v.'s have mean equal to the sum of each individual mean, then that would be much appreciated.

Are you allowed to use the result that the moment generating function of a poission distribution with parameter ##\lambda## is ##M(t) = e^{\lambda (e^t -1)}## ?
 
Stephen Tashi said:
Are you allowed to use the result that the moment generating function of a poission distribution with parameter ##\lambda## is ##M(t) = e^{\lambda (e^t -1)}## ?

That should be ##e^{\lambda (t-1)}.##
 
Stephen Tashi said:

Sorry, no you are correct. I meant the moment-generating function of the probability mass function, while you meant the moment-generating function of the random variable. Of course, they are different. (Your terminology "generating function of a Poisson distribution" threw me: I have seen it used both ways in different sources.)

See, eg., https://web.ma.utexas.edu/users/gordanz/notes/lecture5.pdf
 
Last edited:
It wasnt clear to me what ##s## is in the original post, though I now believe we're classically talking about ##s \in (-1,1)## -- though ##s \in (0,1) ## really is what is of interest -- as the original post appears to already using an Ordinary Generating Function, and hence the identity to be proven comes from the fact that by stochastic independence:

##\text{left hand side} = E\big[s^{X_1}\big]E\big[s^{X_2}\big] = E\big[s^{X_1}s^ {X_2}\big] = E\big[s^{X_1 + X_2}\big] = \text{right hand side}##

OP just needs to confirm that ##g(X) = s^{X}## is a random variable and that the transform doesn't change dependencies (the fact that generating functions are in principle invertible implies this)

- - - -
Equivalently, OP's question seems to be (while using OGFs) that the convolution of two Poissons with parameters ##\lambda ## and ##\mu## is a Poisson with parameter ##\lambda## and ##\mu##. There's a very elegant and probabilistic argument for this that uses memorylessness and the fact that there must be some constant ##\alpha \gt 0## where ##\mu \cdot t = (\alpha \lambda) \cdot t = \lambda \cdot (\alpha t)## ...

- - - -
a less probabilistic take would be to consider properties of the exponential function and simplify. E.g. for starters

##\sum_{n=0}^\infty s^n e^{-\lambda} \frac{\lambda^n}{n!} = e^{-\lambda} \big(\sum_{n=0}^\infty \frac{(s\lambda)^n}{n!}\big)=e^{-\lambda}\big(e^{s\lambda}\big) = e^{-\lambda + s\lambda}##

and apply this process to other parts of the original equation, then simplify.
 
Last edited:
Ray Vickson said:
Does your formula ##(x+y)^n=\sum_{k=0}^n x^ky^{n-k}## work for n=2n = 2 or n=3?

Oops, it should be ##(x+y)^n=\sum_{k=0}^n \binom n k x^ky^{n-k}##.

StoneTemplePython said:
a less probabilistic take would be to consider properties of the exponential function and simplify. E.g. for starters

##\sum_{n=0}^\infty s^n e^{-\lambda} \frac{\lambda^n}{n!} = e^{-\lambda} \big(\sum_{n=0}^\infty \frac{(s\lambda)^n}{n!}\big)=e^{-\lambda}\big(e^{s\lambda}\big) = e^{-\lambda + s\lambda}##

and apply this process to other parts of the original equation, then simplify.

Oh, so ##G_{X+Y}(s)=G_X(s)G_Y(s)=(e^{-\lambda + s\lambda})(e^{-\mu + s\mu})=e^{(\lambda+\mu)(s-1)}=\sum_{n=0}^\infty s^n e^{-(\lambda+\mu)} \frac{(\lambda+\mu)^n}{n!}## implies that ##X+Y## has a distribution ##\text{Poiss}(\lambda+\mu)##?
 
  • Like
Likes StoneTemplePython
  • #10
Eclair_de_XII said:
Oh, so ##G_{X+Y}(s)=G_X(s)G_Y(s)=(e^{-\lambda + s\lambda})(e^{-\mu + s\mu})=e^{(\lambda+\mu)(s-1)}=\sum_{n=0}^\infty s^n e^{-(\lambda+\mu)} \frac{(\lambda+\mu)^n}{n!}## implies that ##X+Y## has a distribution ##\text{Poiss}(\lambda+\mu)##?

Yes. That's really all there is to it from the OGF standpoint. Since an OGF uses a power series in ##s## and the Poisson uses the power series for the exponential function, it should be an easy result.
 
  • #11
Thank you, everyone.
 
Back
Top