ABSTRACT
In this chapter, we treat exclusively estimation problems for the unknown regression parameters in a linear model. We will assume that the errors are normally distributed. We should make it clear that relevant literature in this field and those cited in Chapter 11 have many commonalities. Thus, we may refrain from citing many sources which were included previously. Let us begin with a linear model which is also known as the Gauss-
Markov setup. Suppose that we observe a sequence of independent observations Y1, Y2, . . ., Yn, . . ., referred to as a response or dependent variable. Additionally, we have p covariates X1,X2, . . . ,Xp, which when fixed at certain levels
X1 = xi1,X2 = xi2, . . . ,Xp = xip,
give rise to the ith response Yi. The fixed values xi1, xi2, . . . , xip consti-
tute the ith design point which supposedly leads to a response Yi, i = 1, . . . , n, . . ..
