9.3 Suppose ECMis used to find the ML estimate of in Example 8.6. Further suppose that

9.3 Suppose ECMis used to find the ML estimate of θ in Example 8.6. Further suppose that SEM is applied to the sequence of ECM iterates, assuming they were EMiterates. Are the iterates likely to converge more quickly or more slowly than EM?Would the resulting asymptotic covariancematrix more likely be an over or underestimate, and explain your reasoning.

Comment on the possible asymmetry of the calculated matrix? (See Van Dyk et al. 1995, for details.)