Difficulties sometimes present themselves when statisticians are faced with the task of making accurate predictions from sets of data which have gaps – in the sense that some of the data are missing. The question then arises – are the missing data ignorable or non-ignorable?
For a background on how the complexities of Bayesian mathematics can help towards resolving these conundra, see this .pdf of a slide presentation based on the paper ‘Non-ignorable missingness‘ by Michel Mouchart, Professeur émérite at the Université catholique de Louvain, Institut de statistique (UCL/STAT), Belgium.
The paper “Makes the concept of ‘missingness’ precise.”