What are latest statistical challenges regarding missing data? Could you suggest more recent reviews or book on the methods for handling missing data (i.e. multiple imputation)?
I want to know the challenges in this field.
 A: If you are a biostat student in a top program looking for a good technical challenge for a dissertation (and I am at a loss imagining any other situation in which a person would ask a question like this on CV, rather than browse the recent issues of JASA and Biometrika to find out what people are concerned with), you can start stealing from econometrics: they've had a concept of partial identification of probability distributions in presence of missing data, and I have not seen it propagated into statistics yet. (The person who reworked the idea of the generalized method of moments for biostatisticians got a very nice career out of it.)
A: The canonical reference is Rubin - Missing Data Analysis.
Apart from that, I recently read a paper on diagnostics in multiple imputation by Gelman and others that seemed to point towards a number of different imputation methods (new to me, but I'm not a statistician). The link above requires a subscription, but there's a free copy available from Gelman's website and through Google Scholar.
A: I think you might want to take a look at the recent report (PDF preview) on missing data from the FDA.
