8.16. Suppose that (a) X is Bernoulli with PrX 1 1 PrX 0

8.16. Suppose that

(a) X is Bernoulli with PrðX ¼ 1Þ ¼ 1  PrðX ¼ 0Þ ¼ p, and

(b) Y given X ¼ j is normal with mean mj, variance s2, a simple form of the discriminant analysis model. Consider now the monotone missing-data pattern with Y completely observed but n  r values of X missing, and an ignorable mechanism. In Problem 7.17 we found that the factored likelihood method of Chapter 7 does not provide closed-form expressions for ML estimates. Describe the E and M steps of the EM algorithm for this problem, and provide a flow chart for programming this algorithm.