1

MASSACHUSETTSINSTITUTEOFTECHNOLOGY
6.436J/15.085J Fall 2008
Lecture 1 9/3/2008
PROBABILISTICMODELSANDPROBABILITYMEASURES
Contents
1. Probabilistic experiments
2. Sample space
3. Discrete probability spaces
4. -elds
5. Probability measures
6. Continuity of probabilities
PROBABILISTICEXPERIMENTS
Probability theory is a mathematical framework that allows us to reason about
phenomena or experiments whose outcome is uncertain. A probabilistic model
is a mathematical model of a probabilistic experiment that satises certain math-
ematical properties (the axioms of probability theory), and which allows us to
calculate probabilities and to reason about the likely outcomes of the experi-
ment.
A probabilistic model is dened formally by a triple (,F,P), called a
probabilityspace, comprised of the following three elements:
(a) is the samplespace, the set of possible outcomes of the experiment.
(b) F is a -eld, a collection of subsets of .
(c) Pis a probabilitymeasure, a function that assigns a nonnegative probabil-
ity to every set in the -eld F.
Our objective is to describe the three elements of a probability space, and
explore some of their properties.
1
2 SAMPLESPACE
The sample space is a set comprised of all the possible outcomes of the ex-
periment. Typical elements of are often denoted by , and are called ele-
mentaryoutcomes, or simply outcomes. The sample space can be nite, e.g.,
= {
1
, . . . ,
n
}, countable, e.g., = N, or uncountable, e.g., = R or
={0,1}

.
As a practical matter, the elements of must be mutually exclusive and
collectively exhaustive, in the sense that once the experiment is carried out, there
is exactly one element of that occurs.
Examples
(a) If the experiment consists of a single roll of an ordinary die, the natural sample
space is the set = {1,2, . . . ,6}, consisting of 6 elements. The outcome = 2
indicates that the result of the roll was 2.
(b) If the experiment consists of ve consecutive rolls of an ordinary die, the natural
sample space is the set = {1,2, . . . ,6}
5
. The element = (3,1,1,2,5)is an
example of a possible outcome.
(c) If the experiment consists of an innite number of consecutive rolls of an ordinary
die, the natural sample space is the set ={1,2, . . . ,6}

. In this case, an elemen-

tary outcome is an innite sequence, e.g., =(3,1,1,5, . . .). Such a sample space
would be appropriate if we intend to roll a die indenitely and we are interested in
studying, say, the number of rolls until a 4is obtained for the rst time.
(d) If the experiment consists of measuring the velocity of a vehicle with innite preci-
sion, a natural sample space is the set Rof real numbers.
Note that there is no discussion of probabilities so far. The set simply
species the possible outcomes.
3 DISCRETEPROBABILITYSPACES
Before continuing with the discussion of -elds and probability measures in
their full generality, it is helpful to consider the simpler case where the sample
space is nite or countable.
2

Denition1. A discreteprobabilityspaceis a triplet (,F,P)such that:

(a) The sample space is nite or countable: ={
1
,
2
, . . .}.
(b) The -eld F is the set of all subsets of .
(c) The probability measure assigns a number in the set [0,1] to every subset
of . It is dened in terms of the probabilities P({}) of the elementary
outcomes, and satises
P(A) = P({}),
A
for every A, and

P({}) = 1.

For simplicity, we will usually employ the notation P()instead of P({}),

and we will often denote P(
i
)by p
i
.
The following are some examples of discrete probability spaces. Note that
typically we do not provide an explicit expression for P(A)for every A. It
sufces to specify the probability of elementary outcomes, from which P(A)is
readily obtained for any A.
Examples.
(a) Consider a single toss of a coin. If we believe that heads (H) and tails (T) are equally
likely, the following is an appropriate model. We set ={
1
,
2
), where
1
=H
and
2
= T, and let p
1
= p
2
= 1/2. Here, F = {,{H},{T},{H, T}}, and
P() = 0, P(H) =P(T) = 1/2, P({H, T}) = 1.
(b) Consider a single roll of a die. if we believe that all six outcomes are equally likely,
the following is an appropriate model. We set = {1,2, . . . ,6}and p
1
= =
p
6
= 1/6.
(c) This example is not necessarily motivated by a meaningful experiment, yet it is a
legitimate discrete probability space. Let ={1,2,5, a, v, aaa,}, and P(1)=.1,
P(2)=.1, P(5)=.3, P(a) =.15, P(v) =.15, P(aaa) =.2, P() = 0.
(d) Let =N, and p
k
=(1/2)
k
, for k= 1,2, . . . . More generally, given a parameter
p [0,1), we can dene p
k
= (1p)p
k1
, for k = 1,2, . . . . This results in a
legitimate probability space because

(1p)p
k1
= 1.
k=1
(e) Let = N. We x a parameter > 0, and let p
k
= e

k
/k!, for every k N.
This results in a legitimate probability space because

k=1
e

k
/k! = 1.
(f) We toss an unbiased coin ntimes. We let ={0,1}
n
, and if we believe that all se-
quences of heads and tails are equally likely, we let P() = 1/2
n
for every .
3

4
(g) We roll a die ntimes. We let ={1,2, . . . ,6}
n
, and if we believe that all elemen-
tary outcomes (6-long sequences) are equally likely, we let P() = 1/6
n
for every
.
Given the probabilities p
i
, the problem of determining P(A)for some sub-
set of is conceptually straightforward. However, the calculations involved in
determining the value of the sum
A
P()can range from straightforward to
daunting. Various methods that can simplify such calculations will be explored
in future lectures.
-FIELDS
When the sample space is uncountable, the idea of dening the probability of
a general subset of in terms of the probabilities of elementary outcomes runs
into difculties. Suppose, for example, that the experiment consists of drawing
a number from the interval [0,1], and that we wish to model a situation where all
elementary outcomes are equally likely. If we were to assign a probability of
zero to every , this alone would not be of much help in determining the proba-
bility of a subset such as [1/2,3/4]. If we were to assign the same positive value
to every , we would obtain P({1,1/2,1/3, . . .}) = , which is undesirable.
A way out of this difculty is to work directly with the probabilities of more
general subsets of (not just subsets consisting of a single element).
Ideally, we would like to specify the probability P(A) of every subset of
. However, if we wish our probabilities to have certain intuitive mathematical
properties, we run into some insurmountable mathematical difculties. A so-
lution is provided by the following compromise: assign probabilities to only a
partial collection of subsets of . The sets in this collection are to be thought
of as the nice subsets of , or, alternatively, as the subsets of of interest.
Mathematically, we will require this collection to be a -eld, a term that we
dene next.
4
Denition2. Given a sample space , a -eldis a collection F of subsets
of , with the following properties:
(a) F.
(b) If A F, then A
c
F.
(c) If A
i
F for every iN, then

i=1
A
i
F.
Aset Athat belongs to Fis called an event, an F-measurableset, or simply
a measurableset. The pair (,F)is called a measurablespace.
The term event is to be understood as follows. Once the experiment is
concluded, the realized outcome either belongs to A, in which case we say
that the event Ahas occurred, or it doesnt, in which case we say that the event
did not occur.
Exercise1.
(a) Let F be a -eld. Prove that if A, B F, then AB F. More generally, given
a countably innite sequence of events A
i
F, prove that

i=1
A
i
F.
(b) Prove that property (a) of -elds (that is, F) can be derived from properties
(b) and (c), assuming that the -eld F is non-empty.
The following are some examples of -elds. (Check that this is indeed the
case.)
Examples.
(a) The trivial -eld, F ={,}.
(b) The collection F ={, A, A
c
,}, where Ais a xed subset of .
(c) The set of all subsets of : F = 2

|A}. ={A
(d) Let = {1,2, . . . ,6}
n
, the sample space associated with n rolls of a die. Let
A = { = (
1
, . . .
n
) |
1
2}, B = { = (
1
, . . . ,
n
) | 3
1
4}, and
C ={= (
1
, . . . ,
n
)|
1
5}, and F ={, A, B, C, AB, AC, BC,}.
Example (d) above can be thought of as follows. We start with a number
of subsets of that we wish to have included in a -eld. We then include
more subsets, as needed, until a -eld is constructed. More generally, given a
collection of subsets of , we can contemplate forming complements, countable
unions, and countable intersections of these subsets, to form a new collection.
We continue this process until no more sets are included in the collection, at
which point we obtain a -eld. This process is hard to formalize in a rigorous
manner. An alternative way of dening this -eld is provided below. We will
need the following fact.
5
5
Proposition1. Let Sbe an index set (possibly innite, or even uncountable),
and suppose that for every s we have a -eld F
s
of subsets of the same
sample space. Let F = , i.e., a set A belongs to F if and only if
sS
F
s
A F
s
for every sS. Then F is a -eld.
Proof. We need to verify that F has the three required properties. Since each
F
s
is a -eld, we have F
s
, for every s, which implies that F. To
establish the second property, suppose that A F. Then, A F
s
, for every s.
Since each sis a -eld, we have A
c
F
s
, for every s. Therefore, A
c
F,
as desired. Finally, to establish the third property, consider a sequence {A
i
}of
elements of F. In particular, for a given sS, every set A
i
belongs to F
s
. Since
is a -eld, it follows that

. Since this is true for every sS, F

s
i=1
A
i
F
s
it follows that

i=1
A
i
F
s
. This veries the third property and establishes that
F is indeed a -eld.
Suppose now that we start with a collection C of subsets of , which is not
necessarily a -eld. We wish to form a -eld that contains C. This is always
possible, a simple choice being to just let F = 2

. However, for technical

reasons, we may wish the -eld to contain no more sets than necessary. This
leads us to dene F as the intersection of all -elds that contain C. Note that
if His any other -eld that contains C, then F H. (This is because F was
dened as the intersection of various -elds, one of which is H.) In this sense,
Fis the smallest-eld containing C. The -eld Fconstructed in this manner
is called the -eld generatedby C, and is often denoted by (C).
Example.Let =[0,1]. The smallest -eld that includes every interval [a, b][0,1]
is hard to describe explicitly (it includes fairly complicated sets), but is still well-dened,
by the above discussion. It is called the Borel -eld, and is denoted by B. A set
A[0,1]that belongs to this -eld is called a Borelset.
PROBABILITYMEASURES
We are now ready to discuss the assignment of probabilities to events. We have
already seen that when the sample space is countable, this can be accom-
plished by assigning probabilities to individual elements . However, as
discussed before, this does not work when is uncountable. We are then led
to assign probabilities to certain subsets of , specically to the elements of a
-eld F, and require that these probabilities have certain natural properties.
Besides probability measures, it is convenient to dene the notion of a mea-
sure more generally.
6

Denition3. Let (,F)be a measurable space. A measureis a function

:F [0,], which assigns a nonnegative extended real number (A)to
every set Ain F, and which satises the following two conditions:
(a) () = 0;
(b) (Countableadditivity)If {A
i
}is a sequence of disjoint sets that belong to
F, then (
i
A
i
) =

i=1
(A
i
).
A probabilitymeasureis a measure Pwith the additional property P()=
1. In that case, the triple (,F,P)is called a probabilityspace.
For any A F, P(A)is called the probability of the event A. The assign-
ment of unit probability to the event expresses our certainty that the outcome
of the experiment, no matter what it is, will be an element of . Similarly, the
outcome cannot be an element of the empty set; thus, the empty set cannot occur
and is assigned zero probability. If an event A F satises P(A) = 1, we say
that Aoccurs almostsurely. Note, however, that Ahappening almost surely is
not the same as the condition A = . For a trivial example, let = {1,2,3},
p
1
=.5, p
2
=.5, p
3
= 0. Then the event A={1,2}occurs almost surely, since
P(A) =.5 +.5 = 1, but A= . The outcome 3 has zero probability, but is still
possible.
The countable additivity property is very important. Its intuitive meaning
is the following. If we have several events A
1
, A
2
, . . ., out of which at most
one can occur, then the probability that one of them will occur is equal to
the sum of their individual probabilities. In this sense, probabilities (and more
generally, measures) behave like the familiar notions of area or volume: the area
or volume of a countable union of disjoint sets is the sumof their individual areas
or volumes. Indeed, a measure is to be understood as some generalized notion
of a volume. In this light, allowing the measure (A)of a set to be innite is
natural, since one can easily think of sets with innite volume.
The properties of probability measures that are required by Denition 3 are
often called the axioms of probability theory. Starting from these axioms, many
other properties can be derived, as in the next proposition.
7

Proposition2. Probability measures have the following properties.

(a) (Finiteadditivity)If the events A
1
, . . . , A
n
are disjoint, then P(
n
A
i
) =

i=1
n
P(A
i
).
i=1
(b) For any event A, we have P(A
c
) = 1P(A).
(c) If the events Aand Bsatisfy AB, then P(A)P(B).
(d) (Unionbound)For any sequence {A
i
}of events, we have

P A
i
P(A
i
).
i=1 i=1
(e) (Inclusion-exclusion formula)For any collection of events A
1
, . . . , A
n
,

n
P A
i
= P(A
i
) P(A
i
A
j
)
i=1 i=1 (i,j): i<j
+ P(A
i
A
j
A
k
) + + (1)
n
P(A
1
A
n
).
(i,j,k): i<j<k
Proof.
(a) This property is almost identical to condition (b) in the denition of a mea-
sure, except that it deals with a nite instead of a countably innite collec-
tion of events. Given a nite collection of disjoint events A
1
, . . . , A
n
, let us
dene A
k
= for k > n, to obtain an innite sequence of disjoint events.
Then,
n n

P A
i
=P A
i
= P(A
i
) = P(A
i
).
i=1 i=1 i=1 i=1
Countable additivity was used in the second equality, and the fact P() = 0
was used in the last equality.
(b) The events A and A
c
are disjoint. Using part (a), we have P(AA
c
) =
P(A) +P(A
c
). But AA
c
= , whose measure is equal to one, and the
result follows.
(c) The events Aand B\Aare disjoint. Also, A(B\A) = B. Therefore,
using also part (a), we obtain P(A)P(A) +P(B\A) =P(B).
(d) Left as an exercise.
8

(e) Left as an exercise; a simple proof will be provided later, using random
variables.
Part (e) of Proposition 2 admits a rather simple proof that relies on random
variables and their expectations, a topic to be visited later on. For the special
case where n= 2, the formula simplies to
P(AB) =P(A) =P(B)P(AB).
Let us note that all properties (a), (c), and (d) in Proposition 2 are also valid
for general measures (the proof is the same). Let us also note that for a proba-
bility measure, the property P() = 0need not be assumed, but can be derived
from the other properties. Indeed, consider a sequence of sets A
i
, each of which
is equal to the empty set. These sets are disjoint, since = . Applying
the countable additivity property, we obtain

P() = P() P() = 1,

i=1
which can only hold if P() = 0.
FiniteAdditivity
Our denitions of -elds and of probability measures involve countable unions
and a countable additivity property. A different mathematical structure is ob-
tained if we replace countable unions and sums by nite ones. This leads us to
the following denitions.
Denition4. Let be a sample space.
(a) A eldis a collection F
0
of subsets of , with the following properties:
(i) F.
(ii) If A F, then A
c
F.
(iii) If A F and B F, then AB F.
(b) Let F
0
be a eld of subsets of . A function P : F
0
[0,1]is said to be
nitelyadditiveif
A, B F
0
, AB = P(AB) =P(A) +P(B).
We note that nite additivity, for the two case of two events, easily implies
nite additivity for a general nite number nof events, namely, the property in
part (a) of Proposition 2. To see this, note that nite additivity for n= 2allows
9
6
us to write, for the case of three disjoint events,
P(A
1
A
2
A
3
) =P(A
1
) +P(A
2
A
3
) =P(A
1
) +P(A
2
) +P(A
3
),
and we can proceed inductively to generalize to the case of nevents.
Finite additivity is strictly weaker than the countable additivity property of
probability measures. In particular, nite additivity on a eld, or even for the
special case of a -eld, does not, in general, imply countable additivity. The
reason for introducing the stronger countable additivity property is that with-
out it, we are severely limited in the types of probability calculations that are
possible. On the other hand, nite additivity is often easier to verify.
CONTINUITYOFPROBABILITIES
Consider a probability space in which = R. The sequence of events A
n
=
[1, n]converges to the event A=[1,), and it is reasonable to expect that the
probability of [1, n]converges to the probability of [1,). Such a property is
established in greater generality in the result that follows. This result provides
us with a few alternative versions of such a continuity property, together with
a converse which states that nite additivity together with continuity implies
countable additivity. This last result is a useful tool that often simplies the
verication of the countable additivity property.
Theorem 1. (Continuity of probability measures) Let F be a -eld of
subsets of , and suppose that P:F [0,1]satises P()=1as well as
the nite additivity property. Then, the following are equivalent:
(a) Pis a probability measure (that is, it also satises countable additivity).
(b) If {A
i
}is an increasing sequence of sets in F (i.e., A
i
A
i+1
, for all i),
and A=

A
i
, then lim
i
P(A
i
) =P(A).
i=1
(c) If {A
i
}is a decreasing sequence of sets in F (i.e., A
i
A
i+1
, for all i),
and A=

A
i
, then lim
i
P(A
i
) =P(A).
i=1
(d) If {A
i
}is a decreasing sequence of sets (i.e., A
i
A
i+1
, for all i) and

i=1
A
i
is empty, then lim
i
P(A
i
) = 0.
Proof. We rst assume that (a) holds and establish (b). Observe that A=A
1

(A
2
\A
1
)(A
3
\A
2
). . ., and that the events A
1
, (A
2
\A
1
), (A
3
\A
2
), . . .
10

are disjoint (check this). Therefore, using countable additivity,

P(A) =P(A
1
) + P(A
i
\A
i1
)
i=2
n
=P(A
1
)+ lim P(A
i
\A
i1
)
n
i=2
n
=P(A
1
)+ lim P(A
i
)P(A
i1
)
n
i=2
=P(A
1
) + lim (P(A
n
)P(A
1
))
n
= lim P(A
n
).
n
Suppose now that property (b) holds, let A
i
be a decreasing sequence of sets,
and let A=

A
i
. Then, the sequence A
c
is increasing, and De Morgans law,
i=1 i
together with property (b) imply that

P(A
c
) =P
i=1
A
i
c
=P
i=1
A
i
c
= lim P(A
i
c
),
n

and
P(A) = 1P(A
c
) = 1 lim P(A
c
) = lim 1P(A
c
) = lim P(A
i
).
i i
n n n
Property (d) follows from property (c), because (d) is just the special case of
(c) in which the set Ais empty.
To complete the proof, we now assume that property (d) holds and establish
that property (a) holds as well. Let B
i
F be disjoint events. Let A
n
=

i=n
B
i
. Note that {A
n
} is a decreasing sequence of events. We claim that

n=1
A
n
=. Indeed, if A
1
, then B
n
for some n, which implies that
/
i=n+1
B
i
= A
n+1
. Therefore, no element of A
1
can belong to all of the

sets A
n
, which means that that the intersection of the sets A
n
is empty. Property
(d) then implies that lim
n
P(A
n
) = 0.
Applying nite additivity to the ndisjoint sets B
1
, B
2
, . . . , B
n1
,

B
i
,
i=n
we have

n1
P B
i
= P(B
i
) +P B
i
.
i=1 i=1 i=n
This equality holds for any n, and we can take the limit as n . The rst term
on the right-hand side converges to

i=1
P(B
i
). The second term is P(A
n
), and
11
as observed before, converges to zero. We conclude that
P

i=1
B
i

=
n

i=1
P(B
i
),
and property (a) holds.
References
(a) Grimmett and Stirzaker, Chapter 1, Sections 1.1-1.3.
(b) Williams, Chapter 1, Sections 1.0-1.5, 1.9-1.10.
12
MIT OpenCourseWare
http://ocw.mit.edu
6.436J / 15.085J Fundamentals of Probability
Fall 2008
For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms.