Strong Law of Large Numbers

Graduate Depth 90 in the knowledge graph I know this Set as goal
Unlocks 38 downstream topics
law-of-large-numbers limit-theorems probability

Core Idea

If {Xₙ} are i.i.d. with finite mean μ, then Sₙ/n converges almost surely to μ: P(lim_{n→∞} Sₙ/n = μ) = 1. This is stronger than the weak law. The proof uses the Borel-Cantelli lemmas (for bounded random variables) or truncation arguments. The SLLN provides certainty (up to sets of probability zero) rather than just high probability.

Explainer

You know from the Weak Law of Large Numbers that for any ε > 0, P(|Sₙ/n − μ| > ε) → 0 as n → ∞. This says that for any fixed threshold, the probability of being far from the mean goes to zero. But it leaves open a disconcerting possibility: the sample average could wander far from μ infinitely often, as long as those excursions become increasingly rare. The Strong Law of Large Numbers closes this gap: with probability 1, the sample average *actually converges* to μ — meaning you could observe the entire infinite sequence X₁, X₂, X₃, … and the running average would settle down permanently to μ, not just occasionally get close.

The difference between weak and strong convergence is precisely the difference you studied between convergence in probability and almost sure convergence. Almost sure convergence requires P({ω : Sₙ(ω)/n → μ}) = 1 — the set of sample paths on which the average fails to converge has probability zero. This is a statement about the whole trajectory, not just about snapshots at individual n. It is possible for Sₙ/n to converge in probability to μ without converging almost surely — but the SLLN guarantees both simultaneously.

The Borel-Cantelli lemmas are the key tools in the proof for bounded random variables. First Borel-Cantelli says: if Σ P(Aₙ) < ∞, then P(infinitely many Aₙ occur) = 0. Applying this to the events Aₙ = {|Sₙ/n − μ| > ε}: the goal is to show the sum of their probabilities converges, which implies the average can exceed ε for only finitely many n (with probability 1). For bounded variables, Chebyshev-like tail bounds give P(Aₙ) ≤ C/n², whose sum converges. For unbounded i.i.d. variables with finite mean, a truncation argument handles the heavy tails separately — approximate the Xᵢ by truncated versions, prove the SLLN for those, then show the truncation error is negligible almost surely.

The practical meaning is profound. If you run a casino game with house edge μ > 0 indefinitely, the SLLN says your profit per game *will* converge to μ — not just with high probability, but with certainty in the measure-theoretic sense. Actuaries rely on this when pricing insurance over large portfolios. Physicists rely on it when equating time averages with ensemble averages in ergodic systems. The SLLN is what transforms μ from a theoretical expectation into an empirically observable frequency — the mathematical foundation for the entire enterprise of statistical estimation from data.

Practice Questions 5 questions

Prerequisite Chain

Counting to 10Counting to 20Understanding ZeroThe Number ZeroCounting to FiveOne-to-One CorrespondenceCombining Small Groups Within 5Addition Within 10Addition Within 20Two-Digit Addition Without RegroupingTwo-Digit Addition with RegroupingAddition Within 100Repeated Addition as MultiplicationMultiplication Facts Within 100Division as Equal SharingDivision as Grouping (Measurement Division)Division: Grouping (Repeated Subtraction) ModelDivision: Fair Sharing ModelDivision as Equal SharingDivision as GroupingBasic Division FactsDivision Facts Within 100Two-Digit by One-Digit DivisionDivision with RemaindersRemainders and Quotients in DivisionDivision Word ProblemsIntroduction to Long DivisionFactors and MultiplesPrime and Composite NumbersEquivalent FractionsRelating Fractions and DecimalsDecimal Place ValueReading and Writing DecimalsComparing and Ordering DecimalsAdding and Subtracting DecimalsMultiplying DecimalsDividing DecimalsDividing FractionsMixed Number ArithmeticOrder of OperationsInteger Order of OperationsVariable ExpressionsCombining Like TermsOne-Step EquationsTwo-Step EquationsSolving Multi-Step EquationsEquations with Variables on Both SidesAngle Pairs: Complementary, Supplementary, and VerticalParallel Lines and TransversalsCorresponding AnglesAlternate Interior AnglesTriangle Angle Sum TheoremExterior Angle TheoremTriangle Inequality TheoremSimilar Triangles: AA SimilaritySimilar Triangles: SSS and SAS SimilarityProportions in Similar TrianglesRight Triangle Trigonometry IntroductionTrigonometric Ratios ReviewRadian MeasureConverting Between Degrees and RadiansThe Unit CircleGraphing Sine and CosineGraphing Tangent and Reciprocal Trigonometric FunctionsDerivatives of Trigonometric FunctionsAntiderivativesIndefinite IntegralsBasic Integration RulesRiemann SumsDefinite Integral DefinitionFundamental Theorem of Calculus Part 1Fundamental Theorem of Calculus Part 2U-SubstitutionPartial Fraction Decomposition for IntegrationImproper Integrals - ConvergenceIntegral TestP-SeriesComparison TestLimit Comparison TestAbsolute vs. Conditional ConvergencePower SeriesTaylor PolynomialsTaylor SeriesMoment Generating FunctionsCharacteristic FunctionsConvergence in DistributionStationary DistributionsConvergence of Markov ChainsConvergence in ProbabilityAlmost Sure ConvergenceStrong Law of Large Numbers

Longest path: 91 steps · 502 total prerequisite topics

Prerequisites (3)

Leads To (2)