In the previous lessons, we built the foundation of measure theory:

• Sets
• Sigma algebras
• Measurable spaces

We now arrive at the central object of measure theory:

The Measure

A measure is the mathematical object that assigns a notion of size to sets.

Length is a measure.

Area is a measure.

Volume is a measure.

Probability is a measure.

In fact, one of the deepest insights in modern mathematics is:

Probability is simply a measure whose total size equals one.

This lesson introduces measures formally and develops the intuition that will later become probability theory.

---

The Problem We Want to Solve

Suppose we have the interval:

$$[0,5]$$

Its length is:

$$5$$

Similarly:

$$[0,10]$$

has length:

$$10$$

We intuitively understand length.

But what about:

• complicated unions of intervals?
• fractal sets?
• infinitely many intervals?
• events in probability?

We need a single mathematical framework that works for all of them.

That framework is the measure.

---

The Goal of a Measure

Given a measurable space:

$$\left(X,\mathcal A\right)$$

we want a function that assigns a nonnegative size to every measurable set.

Symbolically:

$$\mu:\mathcal A\rightarrow[0,\infty]$$

where:

$$\mu(A)$$

represents the size of set:

$$A$$

The symbol:

$$\mu$$

is pronounced:

“mu”

and is the standard notation for a measure.

---

Formal Definition

A measure is a function:

$$\mu:\mathcal A\rightarrow[0,\infty]$$

satisfying three axioms.

These axioms encode everything we mean by “size.”

---

Axiom 1: Non-Negativity

For every measurable set:

$$A\in\mathcal A$$

we require:

$$\mu(A)\ge 0$$

A size cannot be negative.

Examples:

Length:

$$5$$

Area:

$$20$$

Probability:

$$0.8$$

All are nonnegative.

---

Axiom 2: Empty Set Has Measure Zero

The empty set contains nothing.

Therefore:

$$\mu(\emptyset)=0$$

This agrees with intuition.

No points means no size.

Examples:

Length of nothing:

$$0$$

Area of nothing:

$$0$$

Probability of impossible event:

$$0$$

---

Axiom 3: Countable Additivity

This is the most important axiom.

Suppose:

$$A_1,A_2,A_3,\ldots$$

are pairwise disjoint sets.

Disjoint means:

$$A_i\cap A_j=\emptyset$$

for:

$$i\neq j$$

Then:

$$\mu\left(\bigcup_{i=1}^{\infty}A_i\right)=\sum_{i=1}^{\infty}\mu(A_i)$$

This is called:

Countable Additivity

or

Sigma Additivity

This single axiom is the heart of measure theory.

---

Why Countable Additivity Matters

Consider:

$$A=[0,2]$$

and

$$B=[2,5]$$

These intervals do not overlap except at a boundary point.

Their lengths are:

$$\mu(A)=2$$

$$\mu(B)=3$$

Therefore:

$$\mu(A\cup B)=5$$

This agrees with ordinary geometry.

Countable additivity extends this idea to infinitely many sets.

---

Finite Additivity vs Countable Additivity

Finite additivity states:

$$\mu(A\cup B)=\mu(A)+\mu(B)$$

for disjoint sets.

Countable additivity states:

$$\mu\left(\bigcup_{i=1}^{\infty}A_i\right)=\sum_{i=1}^{\infty}\mu(A_i)$$

for infinitely many disjoint sets.

The difference may seem small.

It is actually enormous.

Without countable additivity:

• Lebesgue integration fails
• Probability theory breaks
• Brownian motion cannot be defined
• Bayesian nonparametrics collapses

Modern mathematics depends on countable additivity.

---

Example: Counting Measure

Suppose:

$$X={1,2,3,4,5}$$

Define:

$$\mu(A)=\text{number of elements in }A$$

Examples:

$$\mu({1,2})=2$$

$$\mu({3,4,5})=3$$

$$\mu(X)=5$$

This is called the counting measure.

The size of a set is simply the number of elements it contains.

---

Example: Length Measure

Consider:

$$X=\mathbb R$$

Define:

$$\mu([a,b])=b-a$$

Examples:

$$\mu([0,1])=1$$

$$\mu([2,5])=3$$

This eventually becomes Lebesgue measure.

---

Example: Area Measure

For subsets of:

$$\mathbb R^2$$

we define:

$$\mu(\text{rectangle})=\text{length}\times\text{width}$$

Example:

Rectangle:

$$[0,2]\times[0,3]$$

Area:

$$\mu=6$$

Area is another measure.

---

Example: Volume Measure

For subsets of:

$$\mathbb R^3$$

Volume becomes:

$$\mu=\text{length}\times\text{width}\times\text{height}$$

Again we have a measure.

---

Example: Probability Measure

Suppose we toss a fair coin.

Sample space:

$$\Omega={H,T}$$

Define:

$$P(H)=0.5$$

$$P(T)=0.5$$

Notice:

$$P(\emptyset)=0$$

and

$$P(\Omega)=1$$

Probability satisfies all three measure axioms.

Therefore:

Probability is a measure.

The only extra condition is:

$$P(\Omega)=1$$

---

Measure Space

Once we have a measurable space:

$$\left(X,\mathcal A\right)$$

and a measure:

$$\mu$$

we obtain a measure space:

$$\left(X,\mathcal A,\mu\right)$$

This is the fundamental object studied in measure theory.

---

Probability Space Revisited

A probability space is simply a special measure space:

$$\left(\Omega,\mathcal F,P\right)$$

where:

$$P(\Omega)=1$$

Every probability space is a measure space.

Not every measure space is a probability space.

---

Important Consequences

From the measure axioms we can prove:

If:

$$A\subseteq B$$

then:

$$\mu(A)\le\mu(B)$$

This property is called:

Monotonicity

Larger sets cannot have smaller measure.

---

Continuity of Measures

Suppose:

$$A_1\subseteq A_2\subseteq A_3\subseteq\cdots$$

Then:

$$\mu\left(\bigcup_{n=1}^{\infty}A_n\right)=\lim_{n\to\infty}\mu(A_n)$$

Measures behave nicely with limits.

This property becomes crucial for probability theory.

---

Why Measures Matter for Statistics

Every probability distribution is a measure.

Normal distribution:

$$N(\mu,\sigma^2)$$

Poisson distribution:

$$Poisson(\lambda)$$

Gamma distribution:

$$Gamma(\alpha,\beta)$$

Dirichlet Process:

$$DP(\alpha,G_0)$$

All are ultimately measures.

The language of modern statistics is therefore the language of measure theory.

---

Why Measures Matter for Bayesian Nonparametrics

In ordinary Bayesian statistics:

we place a probability measure on parameters.

In Bayesian nonparametrics:

we place probability measures on spaces of probability measures.

For example:

$$G\sim DP(\alpha,G_0)$$

where:

$$G$$

itself is a probability measure.

Without understanding measures first, this statement appears mysterious.

With measure theory, it becomes natural.

---

The Hierarchy So Far

We have now built:

Sets

$$\longrightarrow$$

Sigma Algebras

$$\longrightarrow$$

Measurable Spaces

$$\longrightarrow$$

Measures

The next major step is understanding the most important measure ever invented:

Lebesgue Measure

which generalizes ordinary length and becomes the foundation of modern integration.

---

What You Should Know After This Lesson

You should now understand:

1. A measure assigns size to measurable sets.
2. A measure is written:

$$\mu:\mathcal A\rightarrow[0,\infty]$$

3. Measures satisfy:
   • non-negativity
   • empty set has measure zero
   • countable additivity
4. Length, area, volume, and probability are all measures.
5. A measure space is:

$$\left(X,\mathcal A,\mu\right)$$

6. A probability space is a measure space with:

$$P(\Omega)=1$$

7. Modern probability theory is built entirely on measures.

---

Preview of Lesson 5

Next we study:

Lebesgue Measure

This is the measure that generalizes ordinary length to extremely complicated subsets of the real line and forms the foundation for Lebesgue integration, modern probability, stochastic processes, and Bayesian nonparametric theory.