John Moriarty: C C C 1 2 3 K 1 K

CLASSICAL PROBABILITY 2008
1. PROBABILITY SPACES, RANDOM VARIABLES

AND THE BOREL-CANTELLI LEMMAS
JOHN MORIARTY
1. probability spaces
Suppose we wish to make a probability model. We should do three
things:
Find a way to represent every possible outcome
Decide which sets of outcomes (events) are interesting
Assign a probability to each event
If we do these things in a mathematically consistent way, we can use
theorems of classical probability to draw conclusions about our model.
Some of these conclusions might be surprising, some might be obvious;
they may be used to save lives (rare event modelling) or used to make
money (financial modelling).
Example Modelling football matches; hedge funds. How might you
model a football match, following the above three steps?
How can we ensure that a probability model is mathematically consistent, so that the theorems of classical probability (eg. law of large
numbers, central limit theorem) apply?
Definition 1.1. (Probability space). A probability space is a triplet
(, F, P ) such that
is a set (referred to as the sample space).
F is a -field on , i.e. a collection of subsets of such that
(i) F
(ii) If A F, then also AC F (here AC is the complement
of A, i.e. AC = { : S6 A}.
(iii) If A1 , A2 , A3 , ... F, then
k=1 Ak F.
Elements of F are referred to as events.
P : F [0, 1] and
(i) P () = 1
(ii) If A1 , A2 , A3 , ... F are disjoint, then
P(
Ak ) =
k=1
X
k=1
P (Ak ).
JOHN MORIARTY
Remark It can be shown that not only the union of a countable family
of events is in F (compare (iii) above), but also the intersection of such
a family has to be in F (compare the Exercise sheet 1).
Example 1. Suppose = {A, B, C, D} and that {{A, B}, {C}} F.
Which other subsets must F contain? 2. Suppose P ({A, B}) = 0.5.
What is P ({C, D})? How much freedom of choice remains in this
probability model?
Example A coin is flipped three times, and the outcome is recorded.
The probability space corresponding to the coin being fair is given
by (, F, P ) where
= {HHH, HHT, HT H, HT T, T HH, T HT, T T H, T T T },
F is the set of all subsets of (thus for example {HHH, HT H, T T T }
is an event), and
|A|
P (A) =
8
for any A F (here |A| = the number of elements in the set A). It is
easily checked that (, F, P ) indeed is a probability space.

Definition 1.2. (Independence). Two events A, B F are independent if
P (A B) = P (A)P (B).
More generally, a collection {Ai : i I} F of events is independent
if
Y
\
P (Ai )
P ( Ai ) =
iJ
iJ
for any finite subset J of I.

Example Some independences are obvious, others are not. Again, a
coin is flipped three times. Suppose we knew that all three flips gave the
same resultthen it might seem obvious that this gives no information
about the result of the first flip, so we conclude that these events are
independent. But consider the following events:
A = {HHH, T T T },
B = {HHH, HHT, HT H, T HH}
and
C = {HHT, HT H, T HH}
(i.e. the three events are all three flips give the same result, at least
two of the flips are heads and precisely two of the flips are heads).
Note that
|A|
P (A) =
= 2/8 = 1/4,
8
P (B) = 1/2
and
P (A B) = P ({HHH}) = 1/8.
LECTURE NOTES 1
Thus P (A B) = P (A)P (B), so the events A and B are independent.

However, P (C) = 3/8 but P (A C) = P () = 0. Consequently,
P (A C) 6= P (A)P (C),
so the events A and C are not independent. What about B and C?
Remark A subtlety: see the Exercise sheet 1 for an example of three
events that are pairwise independent but not independent.
2. Random variables
Definition 2.1. (Random variable). A random variable X is a
function X : R such that { : X() x} F for any real
number x.
Question If X is a random variable on (, F, P ) do the following
events belong to F? (i) { : X() > x} F (ii) { : y
X() x} F.
Associated to each random variable X is a distribution function F .
It is defined as
F (x) = P (X x).
All distribution functions have the following properties:
(a) limx F (x) = 0 and limx F (x) = 1.
(b) If x y, then F (x) F (y).
(c) F is right-continuous, i.e. F (x + ) F (x) as & 0 (this is
one of the problems in Exercise sheet 1).
Actually, the properties (a)(c) characterise all distribution functions
in the sense that F is a distribution function of some random variable
if and only if F satisfy (a)(c) above.
Question Can you draw or write down a function which satisfies (a)
and (b) above, but not (c)?
Example Consider the following game between two players A and B:
A coin is flipped three times, and then player B pays the amount 3k
to A, where k is the total number of heads in the three flips. Then the
gain for player A is modelled by a random variable X given by
27 if = HHH
9 if {HHT, HT H, T HH}
X() =
3 if {HT T, T HT, T T H}
1 if = T T T.
The distribution function F of X is given by
0
if x < 1
1/8
if 1 x < 3
4/8 if 3 x < 9
F (x) =
7/8 if 9 x < 27
1
if x 27
JOHN MORIARTY

Example (Indicator functions) Let A F be an event.
indicator of A is the random variable 1A defined by

1 if A
1A () :=
0 if
/ A.
The distribution function F
F (x) =
The
of 1A is given by
0
if x < 0
1 P (A) if 0 x < 1
1
if x 1
If a random variable X takes either a finite or countable number of

values, then it is said to be a discrete random variable. The random
variables in the two examples above are both discrete, since they can
only take the finitely many values 1,3,9 or 27 and 0 or 1, respectively.
If there exists a function f : R R+ such that the distribution
function F of the random variable X satisfies
Z x
f (u) du,
(1)
F (x) =
then X is said to be a continuous random variable. If such a function

f exists, then it is called the density function of X.
Question (i) Is the indicator function a continuous random variable
under this definition? (Hint: apply the fundamental theorem of calculus to (1).) (ii) Suppose we are given a function f : R R+ , and then
define F by (1). Are conditions (a)(c) above necessarily satisfied?
Example If a random variable has a density function of the form

1
(x )2
f (x) =
exp
2 2
2 2
for some parameters and > 0, then the random variable has a
normal distribution. If = 0 and = 1, then the distribution is the
standard normal distribution.

In this course we will deal repeatedly with sequences of independent
random variables.
Definition 2.2. (Independence). Two random variables X : R
and Y : R are independent if the events {X x} and {Y y}
are independent events for all x, y R, i.e. if
P ({X x} {Y y}) = P ({X x})P ({Y y}).
LECTURE NOTES 1
More generally, a family {Xi : i I} of random variables is independent if

\
Y
P ( {Xj xj }) =
P ({Xj xj })
jJ
jJ
for any finite subset J of I.

Question Suppose that A and B are events. Under what circumstances are the indicator functions 1A and 1B independent?
Remark See Example Sheet 1 for an example of three random variables which are pairwise independent but not independent.
3. The Borel-Cantelli Lemmas
Let {An }
n=1 be a sequence of events, and define the event A by
[
A=
An .
m=1 n=m
One may then show the following result.

A = { : An for infinitely many n}
Result :
This set is also usually denoted {An i.o.}.

Question Can you give an argument to justify this result?
Theorem 3.1. (The Borel-Cantelli Lemmas)
(i) Let {An }
n=1 be a sequence of events. Then
P (An ) < = P (An i.o.) = 0.
n=1
(ii) Let {An }

n=1 be a sequence of independent events. Then
P (An ) = = P (An i.o.) = 1.
n=1
Proof. (i) Note that

A :=
An
m=1 n=m
An
n=m
for any m. Thus

P (A)
P (Am ) 0
n=m
as m if n=1 P (An ) is convergent.

(ii) From Problem 5, Exercise Sheet 1 we know that
AC =
[
m=1 n=m
AC
n.
JOHN MORIARTY
Now, for fixed m we have

r
r
Y
\
\
C
C
P (AC
An ) = lim
P(
An ) = lim P (
n)
r
n=m
=
=
lim
n=m
r
Y
n=m
(1 P (An )) lim
n=m
lim exp{
r
X
r
Y
exp{P (An )}
n=m
P (An )} = 0,
n=m
where we have used Problem 2 from Exercise Sheet 1, independence

and the inequality 1 x exp{x}. Consequently,
\
[
\
C
C
P (A ) = P (
An ) = lim P (
AC
n ) = 0,
m
m=1 n=m
n=m
so
P (A) = 1
which finishes the proof.
(i) above is usually referred to as the First Borel-Cantelli Lemma,

whereas (ii) is referred to as the Second Borel-Cantelli Lemma. Note
the assumption of independence in the Second Borel-Cantelli Lemma
is crucial. Indeed, without this assumption the result is not true in
general.
Example Let E F be an event such that 0 < P (E) < 1. Define the
sequence {An }
n=1 of events by An = E for all n. Then clearly
X
X
P (An ) =
P (E) = ,
n=1
n=1
but
P (An i.o.) = P (E) < 1.
This example shows that the assumption of independence in the Second
Borel-Cantelli Lemma is crucial.
Question Suppose that I play a sequence of lottery games. In the nth
game there are n lottery tickets, each with an equal chance of winning.
I buy one ticket for each lottery. Am I always guaranteed another win
if I continue playing long enough?

John Moriarty: C C C 1 2 3 K 1 K

Uploaded by

Copyright:

Available Formats

John Moriarty: C C C 1 2 3 K 1 K

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

John Moriarty: C C C 1 2 3 K 1 K

Uploaded by

Copyright:

Available Formats

CLASSICAL PROBABILITY 2008

1. PROBABILITY SPACES, RANDOM VARIABLES

for any finite subset J of I.

Thus P (A B) = P (A)P (B), so the events A and B are independent.

If a random variable X takes either a finite or countable number of

then X is said to be a continuous random variable. If such a function

More generally, a family {Xi : i I} of random variables is independent if

for any finite subset J of I.

One may then show the following result.

This set is also usually denoted {An i.o.}.

P (An ) < = P (An i.o.) = 0.

(ii) Let {An }

P (An ) = = P (An i.o.) = 1.

Proof. (i) Note that

for any m. Thus

as m if n=1 P (An ) is convergent.

Now, for fixed m we have

where we have used Problem 2 from Exercise Sheet 1, independence

(i) above is usually referred to as the First Borel-Cantelli Lemma,

You might also like

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.