May 10, 2023, 09:30–10:45
Auditorium A3
Maths Job Market Seminar
Abstract
In this talk, I will introduce the framework called post-model-selection inference, which received a fair amount of attention during the last decade. I will begin by introducing the confidence intervals suggested by Berk et al (2013), for Gaussian linear models. These intervals are intended to cover optimal regression coefficients, which depend on the selected set of variables. I will present a first personal contribution, where we are interested in linear predictors obtained from these regression coefficients, and in corresponding asymptotic results (fixed- and high-dimensional). Then, I will present a second personal contribution which is an extension of the confidence intervals of Berk et al (2013) to non-linear and non-Gaussian settings. The new confidence intervals will then be supported by asymptotic results and by favorable numerical comparisons with other classes of confidence intervals in the literature.