The original paper on Data Augmentation (DA) got me thinking it was time to have a careful look at the original paper on EM. These are both highly cited papers, and Google scholar says the DA paper has been cited 2538 times (a lot!) and the EM paper has been cited 31328 times (is that possible?).
EM (which stands of Expectation-Maximization) is something I have heard about a lot, and, as an algorithm, it is even something I understood when I read a blog post relating it to a clustering algorithm that I now cannot find (maybe on the Geomblog). The arguments from the EM to DA paper, which I read recently, makes me think that there is something more than the algorithm going on here. In this later paper, Tanner and Wong identify the importance of Section 4 of the EM paper, which provides a collection of example applications of their EM algorithm. Is there something about this way of thinking that is even more important than the algorithm itself?