A World Without Assumptions

Larry Wasserman. Dept of Statistics, Carnegie Mellon

Many of the methods we have developed to deal with high dimensional problems are built on a set of strong assumptions. These assumptions are dangerous because: (i) they don't hold, (ii) they are not testable and (iii) they lead to fragile procedures. Fortunately, there are assumption-free methods for high dimensional inference that I will present in this talk.

Joint work with: Jing Lei, Max G'Sell, Ryan Tibshirani and Alessandro Rinaldo.