1
$\begingroup$

I have been working on an exercise from Applied Linear Statistical Models - 5th edition- by Kutner.

The question is asking me to obtain the variance-covariance matrix for a polynomial regression of one predictor variable in terms of the original predictor variables before centering. I know the mechanics of what needs to be done and how to do everything, but there is one step I could not complete. Fortunately I have a solution for the problem. That follows below:

enter image description here

My only question is how to obtain the constant matrix $A$ that they did in the solution? I know that the vector I'm using in finding the covariance matrix for the regression parameters will be in this particular situation $A\mathbf{b'}$. Where $\mathbf{b'}$ is just the vector of parameter estimates from the polynomial regression. And from that it will follow that to get the needed covariance matrix I will use the relationship $$\Sigma_{bb} = A \Sigma_{b'b'}A^{t}$$.

This I understand. I'm just stuck in understanding how the matrix $A$ was obtained and the motivation for this matrix in particular.

Would someone be able to provide some insight?

$\endgroup$
1
$\begingroup$

We have \begin{align}b_0' &= b_0- b_1\bar{X}+b_{11}\bar{X}^2\\ b_1' &= b_1 - 2b_{11}\bar{X}\\ b_{11}' &=b_{11} \end{align}

Let's rewrite them in matrix multiplication form.

$$\begin{bmatrix} b_0' \\ b_1'\\ b_{11}'\end{bmatrix} = \begin{bmatrix} 1 & -\bar{X} & \bar{X}^2\\ 0 & 1 & -2\bar{X}\\ 0 & 0 & 1\end{bmatrix} \begin{bmatrix} b_0 \\ b_1 \\ b_{11}\end{bmatrix}$$

That is how we obtain the matrix $A$.

$\endgroup$
1
  • $\begingroup$ Oh Geez...how did I miss this?.....Time to go review that chapter again....Thanks for the help. $\endgroup$
    – dc3rd
    Jun 24 at 6:08

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.