Comments on: Missing data methods in longitudinal studies: a review

摘要

I appreciate the opportunity to discuss this comprehensive review paper by two distinguished researchers who have made many important contributions to missing data methods. The review is impressive in its scope, the taxonomy presented by the authors is useful, and there is much in the review that I like and agree with. A minor quibble on terminology: I prefer the term "complete-case analysis" to "complete data analysis" for the methods in Sect. 3.1, since the latter is more easily confused with the hypothetical analysis that would be carried out if the complete data were available. No review, even one as wide-ranging as this, can include everything, and I take the liberty of mentioning some recent work with colleagues, with some additional references. I also note one area of divergence concerning sensitivity analysis, though this may be more a matter of emphasis than a major substantive disagreement.