Xiao-Li Meng - Struggling with large p and small n? How about (potentially) infinite p and (essentially) zero n?
This is one of my favorite figures from this earlier related paper: http://www.annualreviews.org/doi/full/10.1146/annurev-statistics-010814-020310
It seems you could relate the resolution parameter \(r\) to the relevance of inference to the subpopulation under study (an individual), and thereby claim a neo-Fisherian justification for your approach!
A related paper, where the idea of MR and ANOVA decomposition were first introduced: http://www.stat.harvard.edu/Faculty_Content/meng/COPSS_50.pdf
Please login to add an answer/comment or follow this question.