Probability Spaces (Measure-Theoretic Definition)

Core Idea

A probability space is a triple (Ω, ℱ, P) where Ω is a sample space, ℱ is a sigma-algebra of events, and P is a probability measure satisfying σ-additivity: P(∪ₙAₙ) = ΣₙP(Aₙ) for disjoint countable unions. This measure-theoretic definition extends the axioms of probability to handle infinite sample spaces. It provides the rigorous foundation for modern probability theory.

How It's Best Learned

Review the axioms of probability first. Then see how sigma-algebras enable handling infinite sample spaces rigorously. Work examples: discrete spaces, ℝ with Borel sets, ℝⁿ.

Common Misconceptions

• Thinking the axioms alone guarantee countable additivity; countable additivity must be stated explicitly. - Confusing the sample space with the event space; ℱ ⊆ P(Ω). - Assuming any partition of Ω generates a sigma-algebra.

Explainer

You have worked with the probability axioms — probabilities are non-negative, the total probability is 1, and probabilities of disjoint events add. These axioms work well for finite or countably infinite sample spaces. But for a continuous sample space like a randomly chosen real number in [0, 1], new problems arise: there are uncountably many outcomes, single points have probability zero, and naive notions of "event" run into paradoxes (not all subsets of ℝ can be consistently assigned probabilities). The measure-theoretic framework resolves these problems by rebuilding probability on a rigorous foundation. Its central object is the probability space, a triple (Ω, ℱ, P).

The first component, Ω (the sample space), is the set of all possible outcomes. For a coin flip, Ω = {H, T}. For a random real number, Ω = ℝ. For a stochastic process running over time, Ω might be the set of all continuous paths — an infinite-dimensional space. The second component, ℱ (the sigma-algebra of events), specifies which subsets of Ω are legitimate events — subsets to which P can be consistently assigned a probability. Not every subset can be measured (non-measurable sets exist, by constructions like Vitali sets), so ℱ is a carefully chosen collection that is closed under complementation and countable unions. The standard choice for ℝ is the Borel sigma-algebra, generated by all open intervals.

The third component, P (the probability measure), assigns numbers in [0,1] to events in ℱ with P(Ω) = 1. The crucial axiom is countable additivity (σ-additivity): for any countable collection of pairwise disjoint events A₁, A₂, …, we have P(∪ₙAₙ) = ΣₙP(Aₙ). Finite additivity — what the elementary axioms guarantee — is insufficient for continuous spaces. You cannot compute the probability of an interval by summing probabilities of individual points, because there are uncountably many points and each has probability zero. Countable additivity bridges this gap: it is what allows probability to accumulate over limiting processes without contradiction.

The payoff of this framework is that all of modern probability theory rests on a single coherent foundation regardless of the sample space. Random variables become measurable functions from (Ω, ℱ) to (ℝ, Borel sets). Expectation becomes a Lebesgue integral with respect to P, inheriting all the convergence theorems — dominated convergence, monotone convergence — from measure theory. The language applies equally to discrete distributions (where P is a sum), continuous distributions (where P is an integral against a density), and distributions with mixed or exotic structure. Everything you will encounter in advanced probability — conditional expectation, martingales, stochastic processes, limit theorems — is formulated in terms of the probability space triple. It is the grammar of modern probability.