7.7. Show that for the set-up of Problem 7.6, the estimate of b122 obtained by maximizing the complete-data loglikelihood over parameters and missing data is ~ b122 ¼ ^ b122s^ 22=s^*22, where (in the notation of Section 7.2), s^*22 ¼ ^ b2112s^ 11 þ n1 Pr i¼1 ½yi2  y2  ^ b211ðyi1  y1Þ2 : Hence show that b~ 122 is not consistent for b122 unless the fraction of missing data tends to zero as n ! 1. (See Section 6.3; for help see Little and Rubin, 1983b.)