Prediction Intervals in Regression

College Depth 80 in the knowledge graph I know this Set as goal
Unlocks 4 downstream topics
regression prediction inference

Core Idea

A prediction interval estimates where a new individual observation will fall; a confidence interval estimates the mean response. Prediction intervals are wider because they include both uncertainty in estimating the mean and natural variation around the mean.

Explainer

From your work with inference in linear regression, you know that the fitted line ŷ = β̂₀ + β̂₁x is itself uncertain — it's estimated from data, so it wobbles depending on which sample you draw. A confidence interval for the mean response captures exactly this uncertainty: at a given x value, where might the true population mean μ_Y|x lie? That interval shrinks as sample size grows, because with more data the estimated line stabilizes around the truth.

A prediction interval asks a different and harder question: where will the *next single observation* at that x value land? Even if you knew the regression line perfectly — even with infinite data — individual observations would still scatter around it. That scatter is the irreducible noise term ε, with variance σ². A prediction interval must account for *both* sources of uncertainty: the estimation uncertainty in the mean (which goes to zero as n → ∞) and the irreducible observation-to-observation variance (which does not).

Mathematically, the prediction interval at a given x* is ŷ* ± t* · SE_pred, where SE_pred² = s²(1 + h), with h capturing the leverage of x* and the "1" term being the irreducible variance contribution. The "1 +" is the essential difference: the confidence interval uses SE² = s² · h alone, without the leading 1. Because SE_pred > SE_mean always, prediction intervals are always wider — often substantially so, especially for small samples.

The practical lesson is to match the interval to the question. If you want to know the expected height of all 40-year-old men in a population, use a confidence interval for the mean. If you want to know where one specific 40-year-old man's height will fall, use a prediction interval. Confusing them leads to either false precision (using a confidence interval when you need a prediction interval) or unnecessary alarm (the opposite direction). The confidence interval tells you about the center of a distribution; the prediction interval tells you about the distribution itself.

As x* moves away from x̄ (the center of your data), both interval types widen — leverage h increases the farther you extrapolate. But the prediction interval widens more slowly in relative terms because the "1" dominates when h is small. Near the center of the data, prediction intervals are roughly twice as wide as confidence intervals; far into extrapolation territory, both balloon together. This is why extrapolation with a prediction interval makes caution concrete: you can literally see how much uncertainty you're projecting onto a single future observation.

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 DistributionsCumulative Distribution FunctionsContinuous Random VariablesNormal DistributionCentral Limit TheoremConfidence Intervals for MeansZ-Tests and T-Tests for MeansOne-Sample Z-Test for MeansOne-Sample and Two-Sample T-TestsInference in Linear RegressionPrediction Intervals in Regression

Longest path: 81 steps · 403 total prerequisite topics

Prerequisites (2)

Leads To (1)