O’Bayes15: my tutorial
Here are the slides I made for a short tutorial I will deliver this afternoon for the opening day of the International Workshop on Objective Bayes Methodology, O-Bayes15, held in the city of Valencià, so intricately linked with Bayesians and Bayesianism. The more so as we celebrating this time the career and life of our dear friend Susie. Celebrating with talks and stories, morning runs and afternoon drinks, laughs, tears, and more laughs, even though they cannot equate Susie’s unique and vibrant communicative laugh. I will remember how, at O’Bayes 13, Susie was the one who delivered this tutorial. And how, despite physical frailty and fatigue, she did with her usual energy and mental strength. And obviously again with laugh. I will also remember that the last time I visited Valencià, it was for Anabel Forte’s thesis defence, upon invitation from Susie, and that we had a terrific time, from discussing objective Bayes ideas to eating and drinking local goodies, to walking around the grandiose monuments just built (which presumably contributed to ruin the City of Valencià for quite a while!)
Xi'an's Og 
June 1, 2015 at 5:36 pm
One question I’ve been wondering about recently: supposing one is tempted to perform Bayesian model choice from uninformative, perhaps improper, priors using fractional Bayes factors, is there an adaptive design-based strategy for choosing the size of the training sample?
I understand that a lot of past work has focussed on minimal training samples (e.g. intrinsic Bayes factors), but I would guess that one might be able to make an more strategic decision by starting with a minimal training sample and then progressively adding to the size of the training sample until some kind of utility function tells you, “well, we’ve learnt enough about the parameters of the competing models, it’s now a good idea to go ahead and compute Bayes factors with the remainder of the data”?