Scribd
Upload
25 views

Conditional Expectations E (X - Y) As Random Variables: Sums of Random Number of Random Variables (Random Sums)

1) The document discusses conditional expectations E(X|Y) as random variables. It defines conditional probabilities and expectations, and shows that E(X|Y) is a random variable that depends on Y. 2) It presents two theorems: (i) E[E(X|Y)] = E(X) and (ii) Var(X) = Var[E(X|Y)] + E[Var(X|Y)]. It also discusses sums of random variables and conditional expectations for branching processes. 3) It provides additional properties: if X and Y are independent then E(X|Y) = E(X); E[E(g

Uploaded by

sayed Tamir jan
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
25 views

Conditional Expectations E (X - Y) As Random Variables: Sums of Random Number of Random Variables (Random Sums)

1) The document discusses conditional expectations E(X|Y) as random variables. It defines conditional probabilities and expectations, and shows that E(X|Y) is a random variable that depends on Y. 2) It presents two theorems: (i) E[E(X|Y)] = E(X) and (ii) Var(X) = Var[E(X|Y)] + E[Var(X|Y)]. It also discusses sums of random variables and conditional expectations for branching processes. 3) It provides additional properties: if X and Y are independent then E(X|Y) = E(X); E[E(g

Uploaded by

sayed Tamir jan
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

Probability 2 - Notes 5

Conditional expectations E(X|Y ) as random variables


Conditional expectations were discussed in lectures (see also the second part of Notes 3). The
goal of these notes is to provide a summary of what has been done so far. We start by reminding
the main definitions and by listing several results which were proved in lectures (and Notes 3).

Let X and Y be two discrete r.v.’s with a joint p.m.f. fX,Y (x, y) = P(X = x,Y = y). Remember
that the distributions (or the p.m.f.’s) fX (x) = P(X = x) of X and fY (y) = P(Y = y) of Y are
called the marginal distributions of the pare (X,Y ) and that

fX (x) = ∑ fX,Y (x, y) and fY (y) = ∑ fX,Y (x, y).


y x

def fX,Y (x,y)


If fY (y) 6= 0, the conditional p.m.f. of X|Y = y is given by fX|Y (x|y) = fY (y) and the condi-
tional expectation by
def def
E(X|Y = y) = ∑ x fX|Y (x|y) and, more generally, E(g(X)|Y = y) = ∑ g(x) fX|Y (x|y),
x x

is defined for any real valued function g(X). In particular, E(X 2 |Y = y) is obtained when
g(X) = X 2 and
Var(X|Y = y) = E(X 2 |Y = y) − [E(X|Y = y)]2 .

Remark. We always suppose that ∑x |g(x)| fX|Y (x|y) ≤ ∞.

def
Definition. Denote ϕ(y) = E(X|Y = y). Then E(X|Y ) = ϕ(Y ). In words, E(X|Y ) is a random
variable which is a function of Y taking value E(X|Y = y) when Y = y.

The E(g(X)|Y ) is defined similarly. In particular E(X 2 |Y ) is obtained when g(X) = X 2 and

Var(X|Y ) = E(X 2 |Y ) − [E(X|Y )]2 .

Remark. Note that E(X|Y ) is a random variable whereas E(X|Y = y) is a number (y is fixed).

Theorem 1. (i) E[E(X|Y )] = E(X).


(ii) Var(X) = Var[E(X|Y )] + E[Var(X|Y )].

Proof. See lecture or Notes 3.

Sums of random number of random variables (random sums).

Let X1 , X2 , X3 , .... be a sequence of independent identically distributed random variables (i.i.d.


random variables), each with the same distribution, each having common mean a = E(X) and
variance σ2 = Var(X). Here X is a r.v. having the same distribution as X j . The sum S = ∑Nj=1 X j
where the number in the sum, N is also a random variable and is independent of the X j ’s. The
following statement now follows from Theorem 1.

Theorem 2. (i) E[S] = E(X) × E(N) = aE(N).


(ii) Var(S) = Var(X) × E(N) + [E(X)]2 ×Var(N) = σ2 E(N) + a2Var(N).
Proof. (i) Since E[S|N = n] = E[∑nj=1 X j ] = ∑nj=1 E[X j ] = an, we obtain that E[S|N]] = aN (by
the definition! of E[S|N]]). But then, from (i) of Theorem 1, we obtain that E[S] = E[E[S|N]] =
E[aN] = aE[N]. 2

(ii) Similarly Var(S|N = n) = Var[∑nj=1 X j ] = nσ2 and hence Var(S|N) = σ2 N. By Theorem 1,


(ii) we have that

Var(S) = E[Var(S|N)] +Var(E[S|N]) = E[Nσ2 ] +Var(aN) = σ2 E[N] + a2Var(N).

Example: finding E(Yn ) and Var(Yn ) for a branching process.

Remember that a BP Yn , n = 0, 1, 2, ..., is defined by Y0 = 1 and


(n) (n) (n)
Yn+1 = X1 + X2 + ... + XYn ,

(n)
where r.v.’s X j are independent of each other and have the same distribution as a given integer-
valued r.v. X.

Theorem 2 can be used in order to prove the following statements:

Suppose that E(X) = µ, Var(X) = σ2 . Then:


(i) E(Yn ) = µn
2 n−1 (1−µn )
(ii)If µ 6= 1, then Var(Yn ) = σ µ (1−µ) . If µ = 1 then Var(Yn ) = nσ2 .

Proof. Was given in lectures (and a different proof can be found in Notes 4).

Some additional properties of conditional expectations.

1. If X and Y are independent r.v.’s then E(X|Y ) = E(X).


Proof. As we know, X and Y are independent if and only if fX,Y (x, y) = fX (x) fY (y) or, equiva-
lently, fX|Y (x|y) = fX (x). But then E(X|Y = y) = ∑x x fX|Y (x|y) = ∑x x fX (x) = E(X). 2

2. E[E(g(X)|Y )] = E(g(X))
Proof. Set Z = g(X). Statement (i) of Theorem 1 applies to any two r.v.’s. Hence, applying it
to Z and Y we obtain E[E(Z|Y )] = E(Z) which is the same as E[E(g(X)|Y )] = E(g(X)). 2

This property may seem to be more general statement than (i) in Theorem 1. The proof above
shows that in fact these are equivalent statements.

3. E(XY |Y ) = Y E(X|Y ).
Proof. E(XY |Y = y) = E(yX|Y = y) = yE(X|Y = y) (because y is a constant). Hence, E(XY |Y ) =
Y E(X|Y ) by the definition of the conditional expectation. 2

Corollary. E(XY ) = E[Y E(X|Y )]. Proof. E(XY ) = E[E(XY |Y )] = E[Y E(X|Y )]. 2

Exercise. Use the same method to prove that E(Xh(Y )|Y ) = h(Y )E(X|Y ) for any real valued
function h(y).

You might also like