T-Distribution: Theory and Inference

College Depth 73 in the knowledge graph I know this Set as goal
Unlocks 32 downstream topics
t-distribution

Core Idea

T(k) has heavier tails than N(0,1) and is used when population SD is unknown. Arises when replacing σ with sample s. As k→∞, T(k)→N(0,1). More conservative than z-test, reflecting additional uncertainty from estimating σ.

Explainer

You already know the z-score: if X̄ is a sample mean drawn from a normal population with known σ, then Z = (X̄ − μ)/(σ/√n) follows a standard normal distribution. The z-score is exact. But in practice, σ is almost never known — you must estimate it from the data using the sample standard deviation s. The natural question is: what distribution does (X̄ − μ)/(s/√n) follow? The answer is the t-distribution with k = n−1 degrees of freedom, and understanding why requires seeing what changed when σ was replaced by s.

When you substitute s for σ, you introduce a second source of randomness. The numerator (X̄ − μ) is still random, but now the denominator (s/√n) is also random — s fluctuates across samples. The t-statistic is a ratio of a normal to a scaled chi-squared random variable: formally, T(k) = Z/√(χ²(k)/k) where Z ~ N(0,1) and χ²(k) is independent of Z. The chi-squared distribution in the denominator comes from the fact that (n−1)s²/σ² ~ χ²(n−1). Because you're dividing by something that itself varies, the tails of the resulting distribution are heavier than the standard normal — the occasional small values of the denominator produce large values of the ratio more often than the normal predicts.

The degrees of freedom parameter k controls exactly how heavy those tails are. With k=1 (the Cauchy distribution as a limiting case), the tails are so heavy the mean doesn't even exist. As k increases, the extra uncertainty from estimating σ matters less — after all, with a large sample, s is a very good estimate of σ and its own variability becomes negligible. This is why T(k) → N(0,1) as k→∞: when the denominator's randomness vanishes, the t-statistic is indistinguishable from a z-score. In practice, many textbooks treat k > 30 as "close enough to normal," though the exact boundary depends on how conservative you need to be.

The practical consequence is that t-based inference is always more conservative than z-based inference: t-critical values are larger than z-critical values for the same confidence level, so t-confidence intervals are wider and t-tests require more extreme statistics to reject the null. This conservatism is appropriate — you are acknowledging that you don't know σ precisely. The t-distribution is not a compromise or an approximation to something better; it is the exact correct distribution for this inferential situation, and the fact that it converges to the normal as n grows confirms that the extra conservatism gracefully disappears as the evidence accumulates.

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 DefinitionProbability Density Functions and Continuous DistributionsNormal Distribution: Properties and FundamentalsStandard Normal Distribution and Z-Score StandardizationT-Distribution: Theory and Inference

Longest path: 74 steps · 357 total prerequisite topics

Prerequisites (1)

Leads To (1)