# Homework Help: Characteristic function of Sum of Random Variables

1. Nov 4, 2012

### cutesteph

1. The problem statement, all variables and given/known data
Let X,W,Y be iid with a common geometric density f_x(x)= p(1-p)^x for x nonnegative integer
and p is in the interval (0,1)

What is the characteristic function of A= X-2W+3Y ?

Determine the family of the conditional distribution of X given X+W?

2. Relevant equations
the characteristic function of the geomtric series is p/[1-(1-p)exp(it)]

3. The attempt at a solution
The characteristic function of a sum of random variables is the product of the individual characteristic functions.

So I need to find the characteristic function of X , -2W and 3Y and multiply them together?

2. Nov 4, 2012

### haruspex

Sounds right to me. (It requires them to be independent, which they are.)

3. Nov 5, 2012

### cutesteph

This may sound like a stupid question, but how do I get the density of a discrete random variable when multiplied by a constant? Do I simply subsistute the value in for the variable. Like if I has a binomial random variable X~ Bin (n,p) (I choose n) p^i (1-p)^n-1 for i= 0,1,2,...,n how would I find the density of 5X?

4. Nov 5, 2012

### haruspex

Maybe, but I wouldn't assume that. Start with the def of characteristic function and try to work it out from first principles. Shouldn't be hard.

5. Nov 5, 2012

### Ray Vickson

Of course, discrete random variables do not have densities, but the do have probability mass functions and (cumulative) distribution functions. If p(k), k in K, is a discrete pmf (so that P{X = k} = p(k)) then for Y = 5X, the pmf is P{Y = j} = P{5X=j} = P{X=j/5} = p(j/5) (but only for values of j such that j/5 is in K).

RGV

6. Nov 5, 2012

### cutesteph

So for example if I wanted 2X where X ~Bin(5,1/2) ie (5 choose x)(1/2)^x (1/2)^5-x for x =0,1,2,3,4,5. Then 2X would be Bin(2,1/2) (2 choose x)(1/2)^ (1/2)^2-x for x=0,1,2?

How would I find a joint probably mass function of two binomials? Lets say was was Y= 2X and the other was Z=3x and I want the joint pmf of (Y,Z)?

Last edited: Nov 5, 2012
7. Nov 5, 2012

### Ray Vickson

No. If X ~ Bin(5,1/2) and Y = 2X, we have P{Y = j} = P{X = j/2} = C(5,j/2)/2^5 for j = 0,2,4,6,8,10, and P{Y = j} = 0 for all other j.

RGV

8. Nov 5, 2012

### cutesteph

Also would the characteristic function of A in my orginal question would be phi(t) = [p/(1-(1-p)exp(it))] [p(1-p)/(1-p-exp(it))] [p/(1-exp(it)(1-p)^3] .

9. Nov 5, 2012

### cutesteph

So what happens when I had binomials with different indexes? like X+Y in the quoted example?

10. Nov 5, 2012

### Ray Vickson

You know the basic formulas for probabilities of a sum of independent random variables, and you know the probability mass functions of the individual random variables. Just put it altogether yourself.

RGV

11. Nov 5, 2012

### cutesteph

Got it. Thanks.

Determine the family of the conditional distribution of X given X+W?

X+W is a negative binomial(2,p) since the product of the moment generating function X and W shares the same moment generating function with a binomial (2,P) , which we can do since they are independent and we know that moment generating functions are unique.

We know the negative binomial counts the number of failures proceeding the 2 sucess (in this case) in a sequence of bernoulli trials.

The geometric distribution counts the number of bernoulli trials to get one sucess.

What does it mean by family of the conditional distribution?

12. Nov 5, 2012

### haruspex

I would assume it's asking whether it's binomial or whatever.
It would probably be useful to think about the joint distribution of X and X+W.