← SIG (Susquehanna) Interview Insights
The derivation itself was fine, took the derivative of the residual sum with respect to beta, set it to zero, got to the normal equations, inverted.
Start by deriving the OLS estimator step-by-step: set up the sum of squared residuals, take derivatives with respect to the coefficients, set them to zero, and solve the normal equations. Then clearly state the Gauss-Markov assumptions and explain that under these conditions OLS is BLUE, emphasizing that 'best' means minimum variance among linear unbiased estimators. Finally, clarify that efficiency refers to achieving the lowest variance, and discuss the practical implications.
Pro tip: Connect the Gauss-Markov assumptions to real-world data issues (e.g., heteroskedasticity, autocorrelation) and mention how violations affect inference, showing you understand both theory and practice.
Define the linear regression model and the sum of squared residuals (SSR) as the objective to minimize.
Take partial derivatives of SSR with respect to each coefficient, set them to zero, and solve the resulting normal equations to obtain the OLS formula.
List the classical assumptions: linearity, random sampling, no perfect multicollinearity, zero conditional mean, and homoskedasticity (and no autocorrelation for time series).
Define BLUE: Best Linear Unbiased Estimator, where 'best' means minimum variance. Explain that efficiency means achieving the lowest variance among unbiased linear estimators under the assumptions.
Briefly mention what happens when assumptions are violated (e.g., heteroskedasticity leads to inefficient OLS but still unbiased) and how to address it.
AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.