3 Linear Models

Sections 1 and 2 have been preparation. The analysis of variance models of Section 1 and the regression models of Section 4 are not two subjects: both are \(\underline {Y}=XB+\underline {\varepsilon }\), differing only in what the columns of \(X\) contain — indicators of group membership in one case, measured covariates in the other. This section treats them together.

The programme is: define the general linear model and its assumptions; obtain the least squares estimator and its distribution; establish that it is best among linear unbiased estimators; determine which parameters the data can speak about at all when \(X\) is rank deficient, as every analysis of variance design is; give one test that covers every linear hypothesis; and then apply the whole apparatus to designs with two and three factors.

Questions on this section

Stuck on something here? Ask below and it stays attached to this topic.