Dear all,
the mean squares due to regression in a regression analysis can readily
be expressed as follows:
MSR={1 \over {p-1}} b'(X-\overline{X})'(X-\overline{X})b
where b represents a vector of p-1 least square estimates and X is the n
\times p data matrix and \overline{X} is a n \times p matrix of means.
However, I have some difficulty seeing how to arrive at the well known
expression for the expected value of MSR:
E(MSR)= \sigma^2+{1 \over {p-1}} beta'(X-\overline{X})'(X-\overline{X})beta
with beta=E(b) and \sigma^2 is the variance of the residuals.
Can anyone help me out with this?
With kind regards,
Jerry
|