Department Seminar by Richard Lockhart

Please Note: This seminar will be given online.

Conditional inference after model selection

I will survey the ideas underlying exact conditional inference in regression after model selection.  My focus will be on the ideas: target of inference, interpretation of the probabilistic guarantees, unrealistic assumptions, and computational difficulties.  I won’t, of course, prove anything.  My own involvement in this sort of work is joint with Ryan Tibshirani, Jon Taylor, and Rob Tibshirani.