1. Consider the linear regression models
2. FullModel: Y =X+e;
3. Null Model : Y = e,
4. where X Rnp is of full column rank. Assume that condition (N) holds. Let
5. A Rpp be the identity matrix. Prove that
6. (RSSnull RSSfull)/(dfnull dffull) = (A){A(XX)1A}1A/p,
7. RSSfull /df full 2
8. where and are least squares estimators derived from the full model.