Further Topics in Testing
The theory presented in Chapter 6 regarding testing the general linear hypothesis conﬂicts somewhat with common practice. The ﬁrst principles method for testing presented in Section 6.3 is not the most commonly used. In the discussion of the likelihood ratio test in Section 6.4, the proposed solution to the problem ofminimizing the error sumof squares function Q(b), solving the restricted normal equations (3.10), is not a very popular approach. The two most common methods of hypothesis testing are ﬁtting full and reduced models using reparameterizations or by using type I or sequential sums of squares. These two approaches and two other common testing situations are explored in this chapter.