ABSTRACT
Multiple Imputation (MI), introduced in Chapter 11, is now firmly established as a broadly applicable, practical method for the analysis of partially observed data. While it is probably true that in many missing datasettings there are alternative approaches that can be taken to MI-and these may be more efficient or have a stronger justification in a strictly statistical sense-MI has the advantage of generality and practicality (Carpenter and Kenward, 2013, p. 73).
