Econometrics Exam Cheat Sheet _verified_

: Independent variables aren't perfectly correlated. Exogeneity : (Error term is not correlated with independent variables). Homoscedasticity : Error terms have constant variance. No Autocorrelation : Errors are independent of each other. 3. Hypothesis Testing & Interpretation t-test : . Reject if (usually ~1.96 for 95% confidence). p-value : If (usually 0.05), the result is statistically significant . F-test : Used for joint significance of multiple variables. 4. Violation Diagnostics & Remedies

Box MLR.4 in red. Violations here (omitted variables, measurement error, simultaneity) are the most dangerous. econometrics exam cheat sheet

: Common in time series. Use the Durbin-Watson test. Remedy: Use Cochrane-Orcutt or Newey-West errors. : Independent variables aren't perfectly correlated

| Assumption | Definition (Short) | What happens if violated? | | :--- | :--- | :--- | | | Linear in parameters | Model misspecification | | MLR.2 | Random sampling | Sample selection bias | | MLR.3 | No perfect collinearity | No unique OLS solution | | MLR.4 | Zero conditional mean: E(u|X)=0 | Endogeneity → bias & inconsistency | | MLR.5 | Homoskedasticity: Var(u|X)=σ² | Standard errors wrong (heteroskedasticity) | No Autocorrelation : Errors are independent of each other

How "good" is your model? Use these metrics to evaluate the fit. — total variation in Explained Sum of Squares (SSE): — variation explained by the model. Residual Sum of Squares (SSR): — unexplained variation. R2cap R squared (Coefficient of Determination): . It represents the fraction of variation in explained by Adjusted R2cap R squared : Penalizes the addition of unnecessary variables. 3. The Gauss-Markov Assumptions

$$ \hat\beta_1 = \frac\sum (X_i - \barX)(Y_i - \barY)\sum (X_i - \barX)^2 = \fracCov(X,Y)Var(X) $$