Probability Measures and Random Variables

An intuitive introduction to probability measures and measurable random variables, starting from lengths and areas.

Probability TheoryMeasure TheoryIntermediate

Assigning Numbers to Events in Probability

So far, we’ve learned that in probability theory, the “events” we talk about must belong to:

  • a σ-algebra (to allow set operations), and
  • especially the Borel sets (the natural σ-algebra on ℝ)

The next big question is:

👉 How do we assign actual numbers (probabilities) to those sets?

To answer that, we need the idea of a measure.


What Is a Measure? Start with Length and Area

Although the word “measure” may sound abstract, it’s something we use every day.

For example:

  • The interval [0,3][0, 3] has length 3
  • The set [0,1][2,3][0,1] \cup [2,3] has length 2 (1 + 1)
  • The interval [0,1.5][0,1.5] has length 1.5

→ Measures work not only for continuous intervals but also for disjoint unions of intervals.

For areas:

  • The rectangle [0,2]×[0,5][0,2] \times [0,5] has area 10
  • The triangle {(x,y)0x2,0y52.5x}\{(x,y) \mid 0 \leq x \leq 2, 0 \leq y \leq 5 - 2.5x\} has area 5
  • The unit disk {(x,y)x2+y21}\{(x,y) \mid x^2 + y^2 \leq 1\} has area π\pi

👉 A measure is simply a rule that assigns a size (non-negative number) to each set.

Examples of Length Measures

Assigning lengths to intervals - the most familiar example of measures

[0, 3]

0
1
2
3
4
Length calculation: 3 - 0 = 3

[0,1] ∪ [2,3]

0
1
2
3
4
Interval decomposition:
[0, 1] has length = 1
[2, 3] has length = 1
Total length = 1 + 1 = 2

[0, 1.5]

0
1
2
3
4
Length calculation: 1.5 - 0 = 1.5

📏 Properties of Length Measure

  • • Length of interval [a, b] = b - a
  • • Sum of lengths of disjoint intervals = length of their union
  • • Length of empty set = 0
  • • All lengths are non-negative

What Properties Does a Measure Have?

A measure μ\mu is a function from sets to non-negative numbers satisfying:

  1. Empty set has zero measure

    μ()=0\mu(\varnothing) = 0
  2. Non-negativity

    μ(A)0\mu(A) \geq 0
  3. σ-additivity (countable additivity)
    For disjoint sets A1,A2,A_1, A_2, \dots:

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

This justifies why breaking a shape into parts and adding their areas gives the correct total.


What Is a Probability Measure?

A probability measure is just a measure that totals to 1:

  • P()=0P(\varnothing) = 0
  • P(A)0P(A) \geq 0
  • P(Ai)=P(Ai)P\left(\bigcup A_i\right) = \sum P(A_i)
  • P(Ω)=1P(\Omega) = 1

So it’s just a regular measure that’s normalized to 1.

Example:

  • For a uniform distribution on [0,1][0,1]:
    • The measure is length
    • The set [0,0.3][0,0.3] has length 0.3 → So probability is also 0.3

👉 Probability is just a size ratio in a world where the total size is 1.


Random Variables and Measurability: Why It Matters

A random variable is a function from the sample space to ℝ:

X:ΩRX: \Omega \to \mathbb{R}

But not every function qualifies — it must be measurable.

What is Measurability?

Measurability ensures that “events defined via XX” can be assigned probabilities.

Specifically, it requires:

For any reasonable subset of ℝ (like (,3](-\infty,3]), its preimage under XX must belong to F\mathcal{F} — the collection of sets we can assign probabilities to.

Why is this important?

  • We can only assign probabilities to sets in F\mathcal{F}
  • So anything constructed via XX must land in F\mathcal{F} — and measurability guarantees this

Formal Definition: Measurable Function

A function X:(Ω,F)(R,B)X: (\Omega, \mathcal{F}) \to (\mathbb{R}, \mathcal{B}) is F\mathcal{F}-measurable if:

X1(B)Ffor every BBX^{-1}(B) \in \mathcal{F} \quad \text{for every } B \in \mathcal{B}

Here:

  • B\mathcal{B} is the Borel σ-algebra on ℝ (intervals, open sets, etc.)

Intuition and Examples

  • F\mathcal{F} = the collection of measurable events (can assign probabilities)
  • Measurability = ensures events defined by XX are measurable too

Example 1: Dice Roll

  • Ω={1,2,3,4,5,6},F=2Ω\Omega = \{1,2,3,4,5,6\}, \mathcal{F} = 2^\Omega
  • X(ω)=ωX(\omega) = \omega (the die roll)
  • Event “XX is even” = {2,4,6}F\{2,4,6\} \in \mathcal{F} → measurable!

Example 2: Continuous Uniform

  • Ω=[0,1],F=B([0,1])\Omega = [0,1], \mathcal{F} = \mathcal{B}([0,1])
  • X(ω)=ωX(\omega) = \omega
  • Event “X0.5X \leq 0.5” = [0,0.5]F[0,0.5] \in \mathcal{F} → measurable!

Summary

  • A measure assigns size to sets
  • A probability measure is a measure with total size 1
  • A random variable is a measurable function from Ω\Omega to ℝ
  • Measurability ensures we can assign probabilities to events defined via XX
← Back to Encyclopedia