Sections 1 – 4

Mark all correct answers (multiple answers may be correct)

Why create multiple imputations?

How are MI and MICE related?

What is a full conditional distribution ?

What is/are basic assumption(s) in MI?

Why can we not just use the predicted values (e.g., \(y_{mis} = X_{mis}\beta\)) as imputations?

How can results from analyses performed on multiply imputed data be combined?

Why is it not correct to average the standard errors from multiple analyses of imputed data to obtain the overall results?

Why are iterations necessary in MICE for non-monotone missing patterns?

What does it mean when a chain of (MCMC) samples has converged?

Section 5

Mark all correct answers (multiple answers may be correct)

Good methods to draw imputations:

Bootstrap multiple imputation

In predictive mean matching:

Predictive mean matching