Model Assumptions — 6-Point Checklist
Every OLS regression relies on 6 core assumptions. Violating them reduces reliability:
- Linearity: Relationship is linear → check Partial Regression plots
- Residual normality: Residuals are normally distributed → check Q-Q Plot
- Constant variance: Residual spread is consistent → Breusch-Pagan test
- Independence: No autocorrelation between residuals → Durbin-Watson test
- No multicollinearity: X variables not highly correlated → VIF < 5
- No influential outliers: No single point dominates the model → Cook's D + |Std Residual| > 3
Try it in the app
Try: Run regression → check the Model Health Check card at the top of Results for a live pass/fail summary of all 6 assumptions.