5. In Example 14.3, use an MNAR model to generate missing values in y2, namely Ri Bern(i ), Probit(i ) = 0 + 1Yi2,

5. In Example 14.3, use an MNAR model to generate missing values in y2, namely Ri ∼ Bern(πi ), Probit(πi ) = η0 + η1Yi2, where η0 = 0, η1 = 1. At the imputation stage generate five complete datasets in two ways, first with the MAR MI approach used in Example 14.3, and second using an MNAR MI model

How does using the alternative imputation datasets affect results from the final pooled inference stage?

Yi2 N(aI+PMIY1, 1/TMI). ~ Probit[Pr(R; 1)]=no,MI +1.MIY12.