So given that the least squares estimator of $\beta$ is:

$$ \hat{\beta} = (\mathbf{X}^T \mathbf{X})^{-1}\mathbf{X}^T \mathbf{Y} $$

What is the covariance matrix?

Here is my attempt, I'm not sure what $E[\epsilon \epsilon^T]$ and $E[\epsilon \beta^T]$ are:

enter image description here

Does anyone know how to proceed?


$\beta $ is a constant so $$ E(\epsilon\beta^T) = E(\epsilon)\beta^T .$$

You can compute $E(\epsilon)$ and $E(\epsilon\epsilon^T)$ from your assumptions about the distribution of $\epsilon.$ You didn't specify these here, but usually it is assumed that $\epsilon$ is a vector of independent zero-mean normals all with the same variance $\sigma^2.$ So then $E(\epsilon)$ is the zero vector and $$E(\epsilon\epsilon^T) = E(\epsilon_1^2+\ldots+\epsilon_n^2)$$ where $\epsilon_1\ldots \epsilon_n$ are independent $N(0,\sigma^2).$ (I'm assuming $n$ is the number of data-points.)

I'll leave that for your to compute.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.