Model Specification Testing and Diagnostics

College Depth 80 in the knowledge graph I know this Set as goal
Unlocks 1 downstream topic
model-selection specification testing

Core Idea

Model specification testing evaluates whether chosen functional form, regressor sets, and error structure assumptions are appropriate for the data. Common tests include Ramsey RESET for functional form misspecification and comparison of nested models through F-tests; diagnostic checks examine residuals for deviations from white noise.

Explainer

From your work on multiple regression and hypothesis testing, you know how to estimate a model and test whether individual coefficients are statistically significant. But there is a prior question: is the model itself correctly specified? Significance tests assume the model's functional form is right, the relevant variables are included, and the errors are well-behaved. If those assumptions fail, your t-statistics and F-statistics are meaningless — you are testing hypotheses in a model that misrepresents the data-generating process. Specification testing addresses exactly this: how do we detect when the model is wrong before trusting what it tells us?

The broadest class of specification tests asks whether the functional form is appropriate. The most common is the Ramsey RESET test (Regression Specification Error Test). The logic is elegant: if your linear model is correctly specified, the fitted values Ŷ should already capture all systematic variation in Y, and powers of Ŷ (like Ŷ² and Ŷ³) should have no additional predictive power. The RESET test adds these powers as auxiliary regressors and uses an F-test to check whether they are jointly significant. A rejection is a signal that the original linear model is missing something — possibly a nonlinear relationship, an interaction term, or an omitted variable that enters nonlinearly. What it cannot tell you is *what* is wrong; RESET is a diagnostic, not a prescription.

Testing nested models via F-tests is the second major tool. A restricted model is nested inside an unrestricted model when the restricted model imposes specific parameter constraints (usually setting some coefficients to zero). The F-statistic compares how much explanatory power is lost by imposing the restriction. If the restricted model fits nearly as well — if the loss in R² is small relative to the degrees of freedom consumed — the restriction is not rejected. This framework allows principled comparison of competing specifications that differ in which variables are included.

Residual diagnostics complement formal tests by revealing patterns that indicate model failure. If residuals exhibit heteroskedasticity — variance that changes with fitted values or a regressor — the standard errors are wrong even if the coefficients are unbiased. If residuals are autocorrelated — systematically positive or negative in runs — this often signals a missing dynamic structure. If residuals are non-normal, inference in small samples is unreliable. Plots of residuals against fitted values, against each regressor, and over time (for time-series data) are the first-line tools. Formal tests (Breusch-Pagan for heteroskedasticity, Durbin-Watson for autocorrelation) add statistical precision. Together, specification testing and residual diagnostics form the discipline of checking your model before trusting it.

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-TestsHypothesis Testing in RegressionModel Specification Testing and Diagnostics

Longest path: 81 steps · 418 total prerequisite topics

Prerequisites (2)

Leads To (1)