Study List for Linear Models Exam
This list may not be exhaustive but should be a good start, I encourage you to go through our slides, exercises, and notes from the class to best prepare for the upcoming midterm.
Key Formulas
OLS Estimator:
- \(\hat{\boldsymbol{\beta}} = (\mathbf{X}^T\mathbf{X})^{-1}\mathbf{X}^T\mathbf{y}\)
Normal Equations:
- \(\mathbf{X}^T\mathbf{X}\hat{\boldsymbol{\beta}} = \mathbf{X}^T\mathbf{y}\)
Hat Matrix:
- \(\mathbf{H} = \mathbf{X}(\mathbf{X}^T\mathbf{X})^{-1}\mathbf{X}^T\)
- \(\hat{\mathbf{y}} = \mathbf{H}\mathbf{y}\)
- Properties: \(\mathbf{H}^T = \mathbf{H}\) (symmetric), \(\mathbf{H}^2 = \mathbf{H}\) (idempotent)
Variance of OLS:
- \(\text{Var}(\hat{\boldsymbol{\beta}}) = \sigma^2(\mathbf{X}^T\mathbf{X})^{-1}\)
- \(\text{Var}(\hat{\beta}_j) = \sigma^2[(\mathbf{X}^T\mathbf{X})^{-1}]_{jj}\)
Sums of Squares:
- TSS = \(\mathbf{y}^T\mathbf{y} - n\bar{y}^2\) = \(\sum_{i=1}^n(y_i - \bar{y})^2\)
- SS\(_{\text{Reg}}\) = \(\mathbf{y}^T\mathbf{H}\mathbf{y} - n\bar{y}^2\)
- SSE = \(\mathbf{y}^T(\mathbf{I}-\mathbf{H})\mathbf{y}\)
- Fundamental Identity: TSS = SS\(_{\text{Reg}}\) + SSE
Coefficient of Determination:
- \(R^2 = \frac{\text{SS}_{\text{Reg}}}{\text{TSS}} = 1 - \frac{\text{SSE}}{\text{TSS}}\)
Variance Estimation:
- \(\hat{\sigma}^2 = \frac{\text{SSE}}{n-p}\) (MSE)
Test Statistics:
- t-statistic: \(t = \frac{\hat{\beta}_j}{\text{se}(\hat{\beta}_j)}\) where \(\text{se}(\hat{\beta}_j) = \sqrt{\hat{\sigma}^2[(\mathbf{X}^T\mathbf{X})^{-1}]_{jj}}\)
- F-statistic: \(F = \frac{\text{SS}_{\text{Reg}}/(p-1)}{\text{SSE}/(n-p)} = \frac{\text{MS}_{\text{Reg}}}{\text{MSE}}\)
Random Vector Properties
- \(E[\mathbf{A}\mathbf{Y} + \mathbf{b}] = \mathbf{A}E[\mathbf{Y}] + \mathbf{b}\)
- \(\text{Var}(\mathbf{A}\mathbf{Y} + \mathbf{b}) = \mathbf{A}\text{Var}(\mathbf{Y})\mathbf{A}^T\)
Gauss-Markov Assumptions
- Linearity: \(\mathbf{y} = \mathbf{X}\boldsymbol{\beta} + \boldsymbol{\varepsilon}\)
- Zero mean errors: \(E[\boldsymbol{\varepsilon}] = \mathbf{0}\)
- Constant variance & independence: \(\text{Var}(\boldsymbol{\varepsilon}) = \sigma^2\mathbf{I}\)
- Full rank: \(\mathbf{X}\) has full column rank
Degrees of Freedom
- Error: \(n - p\)
- Regression: \(p - 1\)
- Total: \(n - 1\)
Key Conceptual Points
Orthogonality Condition:
- \(\mathbf{X}^T\hat{\boldsymbol{\varepsilon}} = \mathbf{0}\) (residuals perpendicular to column space)
QR Decomposition:
- \(\mathbf{R}\boldsymbol{\hat\beta} = \mathbf{Q}^T\mathbf{y}\)
- How to backsolve to get \(\hat\beta\)
BLUE:
- Best = minimum variance
- Linear = estimator is linear in \(\mathbf{y}\)
- Unbiased = \(E[\hat{\boldsymbol{\beta}}] = \boldsymbol{\beta}\)
- Estimator
For Unbiased Estimator \(\mathbf{C}\mathbf{y}\):
- Must satisfy: \(\mathbf{C}\mathbf{X} = \mathbf{I}\)
Positive Semi-Definite:
- Matrix \(\mathbf{A}\) is PSD if \(\mathbf{x}^T\mathbf{A}\mathbf{x} \geq 0\) for all vectors \(\mathbf{x}\)
- \(\mathbf{D}\mathbf{D}^T\) is always PSD
Matrix Algebra Rules You’ll Need
- \((\mathbf{AB})^T = \mathbf{B}^T\mathbf{A}^T\)
- \((\mathbf{AB})^{-1} = \mathbf{B}^{-1}\mathbf{A}^{-1}\)
- \((\mathbf{A}^T)^{-1} = (\mathbf{A}^{-1})^T\)