Regarding Example 14.7, show that the REML estimator of satisfies (14.29) (according to the definition under

Regarding Example 14.7, show that the REML estimator of ψ satisfies

(14.29) (according to the definition under non-Gaussian linear mixed models; see Section 12.2), where the fj ’s are the same as in Example 14.6 and a(ψ) is given in Example 14.7. [Hint: You may use the identity V

−1 = V

−1X(X



V

−1X)

−1X



V

−1 + A(A



VA)

−1A



, where V = V (θ) and A is any n × (n − p) matrix of full rank (n is the dimention of y = (yj )1≤j≤m) such that A



X = 0 (e.g., Searle et al. 1992, p.

451).]