Consequences of Violating Assumptions
What You'll Learn
You'll discover what happens to your regression results—coefficients, standard errors, and p-values—when the core assumptions are broken.
Why Violated Assumptions Matter
When you violate the assumptions of linear regression, your model doesn't just become "less accurate"—it produces systematically misleading results. Understanding which assumption breaks matters because each has distinct consequences.
How Each Violation Affects Your Results
Violating Linearity
When the true relationship isn't linear, your coefficient estimates (β₀ and β₁) are biased—they're systematically wrong. Your model captures the average linear trend but misses the actual pattern. Even with infinite data, estimates won't converge to true values.
Violating Independence
Correlated observations (like time series data) cause your standard errors to be underestimated. Your confidence intervals become too narrow and your p-values too small, making you falsely confident about significance. You think relationships are stronger than they are.
Violating Homoscedasticity (Constant Variance)
Heteroscedasticity leaves coefficient estimates unbiased but makes standard errors unreliable. Sometimes they're too small (false significance), sometimes too large (missed real effects). Your hypothesis tests become untrustworthy.
Violating Normality
With large samples, coefficient estimates remain reliable due to the Central Limit Theorem. But in small samples, confidence intervals and p-values become inaccurate—though estimates themselves stay unbiased.
Violating Exogeneity
When X correlates with errors (omitted variable bias, measurement error), your coefficients are biased and inconsistent. This is the most serious violation—no amount of data fixes it. You're measuring the wrong thing entirely.
The Big Picture
Think of assumptions like ingredients in baking: miss one, and the cake might still look okay but taste wrong. Miss another, and it won't rise at all. Different violations corrupt different parts of your inference.
Key Takeaway: Violated assumptions don't just reduce accuracy—they create biased estimates (wrong values), incorrect standard errors (wrong uncertainty), or invalid p-values (wrong conclusions). The type of violation determines which parts of your analysis become unreliable.