18.

175: Lecture 2
Extension theorems: a tool for constructing
measures

Scott Sheeld

MIT

18.175 Lecture 2

1
Outline

Extension theorems

Distributions on R

Extension theorems

2
18.175 Lecture 2
Outline

Extension theorems

Distributions on R

Extension theorems

18.175 Lecture 2
3
Recall the dilemma

Want, a priori, to dene measure of any subsets of [0, 1).

Find that if we allow the axiom of choice and require
measures to be countably additive (as we do) then we run
into trouble. No valid translation invariant way to assign a
nite measure to all subsets of [0, 1).
Could toss out the axiom of choice... but we dont want to.
Instead we will only dene measure for certain measurable
sets. We will construct a -algebra of measurable sets and
let probability measure be function from -algebra to [0, 1].
Price to this decision: for the rest of our lives, whenever we
talk about a measure on any space (a Euclidean space, a
space of dierentiable functions, a space of fractal curves
embedded in a plane, etc.), we have to worry about what the
-algebra might be.
4
18.175 Lecture 2
Recall the dilemma

I On the other hand: always have to ensure that any measure
we produce assigns actual number to every measurable set. A
bigger -algebra means more sets whose measures have to be
dened. So if we want to make it easy to construct measures,
maybe its a good thing if our -algebra doesnt have too
many elements... unless its easier to...

I Come to think of it, how do we dene a measure anyway?

I If the -algebra is something like the Borel -algebra (smallest
-algebra containing all open sets) its a pretty big collection
of sets. How do we go about producing a measure (any
measure) thats dened for every set in this family?

I Answer: use extension theorems.

18.175 Lecture 2
5
Recall denitions

I Probability space is triple (, F, P) where is sample
space, F is set of events (the -algebra) and P : F [0, 1] is
the probability function.

I -algebra is collection of subsets closed under
complementation and countable unions. Call (, F) a
measure space.

I Measure is function : F R satisfying (A) J () = 0
for all A F and countable additivity: (i Ai ) = i (Ai )

for disjoint Ai .

I Measure is probability measure if () = 1.

I The Borel -algebra B on a topological space is the smallest

-algebra containing all open sets.

6
18.175 Lecture 2
Outline

Extension theorems

Distributions on R

Extension theorems

7
18.175 Lecture 2
Outline

Extension theorems

Distributions on R

Extension theorems

8
18.175 Lecture 2
How do we produce measures on R?

I Write F (a) = P (, a] .

I Theorem: for each right continuous, non-decreasing function

F , tending to 0 at and to 1 at , there is a unique

measure dened on the Borel sets of R with

P((a, b]) = F (b) F (a).

I If were given such a function F , then we know how to

compute the measure of any set of the form (a, b].

I We would like to extend the measure dened for these subsets

to a measure dened for the whole algebra generated by

these subsets.

I Seems clear how to dene measure of countable union of

disjoint intervals of the form (a, b] (just using countable

additivity). But are we condent we can extend the denition

to all Borel measurable sets in a consistent way?

9
18.175 Lecture 2
Outline

Extension theorems

Distributions on R

Extension theorems

10
18.175 Lecture 2
Outline

Extension theorems

Distributions on R

Extension theorems

18.175 Lecture 2
11
Algebras and semi-algebras

I algebra: collection A of sets closed under nite unions and
complementation.

I measure on algebra: Have (A) () = 0 for all A in A,

and for disjoint

JA i with union in A we have

(
i=1 A i ) = i=1 (Ai ) (countable additivity).

I Measure on A is -nite if exists countable collection
An A with (An ) < and An = .

I semi-algebra: collection S of sets closed under intersection
and such that S S implies that S c is a nite disjoint union
of sets in S. (Example: empty set plus sets of form
(a1 , b1 ] . . . (ad , bd ] Rd .)

I One lemma: If S is a semialgebra, then the set S of nite
disjoint unions of sets in S is an algebra, called the algebra
generated by S.

18.175 Lecture 2
12
-systems and -systems

I Say collection of sets P is a -system if closed under
intersection.

I Say collection of sets L is a -system if
. L
. If A, B L and A B, then B A L.
. If An L and An A then A L.

I THEOREM: If P is a -system and L is a -system that
contains P, then (P) L, where (A) denotes smallest
-algebra containing A.

18.175 Lecture 2
13
Caratheeodory Extension Theorem

I Theorem: If is a -nite measure on an algebra A then
has a unique extension to the algebra generated by A.

I Detailed proof is somewhat involved, but lets take a look at
it.

I We can use this extension theorem prove existence of a unique
translation invariant measure (Lebesgue measure) on the
Borel sets of Rd that assigns unit mass to a unit cube. (Borel
-algebra Rd is the smallest one containing all open sets of
Rd . Given any space with a topology, we can dene a
-algebra this way.)

14
18.175 Lecture 2
MIT OpenCourseWare
http://ocw.mit.edu

18.175 Theory of Probability

Spring 2014

For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms .
MIT OpenCourseWare
http://ocw.mit.edu

18.175 Theory of Probability

Spring 2014

For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms .