The distribution of the OLS estimator
Summary
- Setup:
- a linear regression model \(y=Xβ+ε\) with a random sample
- \(b={\left( X'X \right)}^{-1}X'y\) is the OLS estimator of \(β\) (both are \(k×1\) )
- \(e=y-Xb\) are the OLS residuals.
- \(s^2=e'e/(n-k)\)
- Assumption: Normally distributed error terms (conditionally on \(x\) )
\[ε_i|x_i \sim N(0,σ^2)\]
- Note: we are making assumptions stronger than the Gauss Markov assumptions.
- Result:
\[b|X \sim N\left( β,σ^2{\left( X'X \right)}^{-1} \right)\]
- Result:
\[ \frac{\left( n-k \right)s^2}{σ^2} \sim χ_{n-k}^2\]
- Result: \(s^2\) is independent of \(b\) .