Normal Linear Regression Model

Graduate Depth 88 in the knowledge graph I know this Set as goal
Unlocks 4 downstream topics
regression normality inference assumptions

Core Idea

The normal regression model assumes u ~ N(0,σ²) in addition to OLS assumptions. This distributional assumption enables hypothesis testing and confidence intervals via t and F statistics, allowing exact inference in finite samples rather than relying on asymptotics.

Explainer

From simple linear regression, you know how to estimate coefficients by minimizing the sum of squared residuals. OLS gives you β̂ with desirable properties under the Gauss-Markov assumptions: it's unbiased and efficient among linear estimators. But those properties say nothing about the *distribution* of β̂ — without knowing the shape of that distribution, you cannot make probability statements about how far your estimate might be from the truth. That's where the normal linear regression model comes in: by adding one additional assumption about the error term's distribution, you unlock the entire apparatus of hypothesis testing and confidence intervals without needing large samples.

The new assumption is that the error terms follow a normal distribution: u ~ N(0, σ²). This is a strong claim — you're asserting not just that errors have mean zero and constant variance (the Gauss-Markov assumptions), but that they're drawn from a bell-shaped distribution. Once you make this assumption, a remarkable thing happens: because OLS is a linear function of the errors, and because linear combinations of normal random variables are also normal, the OLS estimator β̂ is itself normally distributed. Specifically, β̂ ~ N(β, σ²(X'X)⁻¹). You now have the exact sampling distribution of your estimator — not an approximation, but the precise distribution.

This sampling distribution is what makes inference possible. The t-statistic for testing whether a coefficient βⱼ equals some hypothesized value β₀ is (β̂ⱼ - β₀) / se(β̂ⱼ), where the standard error is estimated from the data. Under the null hypothesis and the normality assumption, this statistic follows an exact t-distribution with (n-k) degrees of freedom — where n is sample size and k is the number of parameters including the intercept. Similarly, the F-statistic for testing joint hypotheses (does this entire set of coefficients equal zero?) follows an exact F-distribution under the null. These are finite-sample results: they hold exactly even in small samples, not just approximately as sample size grows.

Without the normality assumption, you can still do inference — but only asymptotically, by appealing to the Central Limit Theorem. As n → ∞, the distribution of β̂ converges to normal regardless of the distribution of the errors (under mild regularity conditions), so t and F tests remain valid approximately in large samples. The practical implication: in small samples, the normality assumption is load-bearing. If you have 15 observations and your errors are highly skewed, your t-test p-values may be meaningfully wrong. In large samples — say, 500+ observations — the asymptotic justification is usually sufficient, and the normal regression model becomes a special case of the more general asymptotic theory rather than a necessary restriction. This is why econometrics courses introduce the normal model first for clean finite-sample theory, then relax it for the large-sample results that dominate applied work.

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-TestsOne-Way ANOVAF-Test and Joint SignificanceR-Squared and Model FitMulticollinearityRobust Standard ErrorsPanel Data: Structure and AdvantagesFixed Effects ModelsDifference-in-DifferencesParallel Trends Assumption: Validity and TestingNormal Linear Regression Model

Longest path: 89 steps · 511 total prerequisite topics

Prerequisites (6)

Leads To (2)