Consider a regression problem with n real-valued response variables y = (y1,..., y,), and a single scalar covariate x; associated with response each variable y, i = 1,..., n. Letting x = (X1, ...,x,), we can assume the covariates in x are not all identical. Notationally, let X;Vi. Xia Sxx Sy Vis Syy Sry Sx 1=1 F1 i=1 Suppose we posit a normal linear model, (1) y x - Normal,(XB, o' I,), where o > 0, B = (81, B2), and X (2) X =VR = ( V -] + XT X ) -', MR = Vn XTy, n an = a+ bn = b+ (y y -yXm.)