Hanna Jankowski (York University)

Math & Stats Department Colloquium

Friday, March 31st, 2023

Time: 2:30 p.m.  Place: Jeffery Hall, Room 234

Speaker: Hanna Jankowski (York University)

Title: Some results for the Isotonic Single Index model

Abstract: The semiparametric single index model is a popular dimension reduction tool used in a variety of applications. It can be viewed as a further generalization of generalized linear models, where the link function is not specified. In this talk, I will discuss the isotonic version of the nonparametric single index model, where the link function is assumed to be increasing. A key benefit of the assumption is that this approach is free of any additional tuning parameters, such as the bandwidth. Furthermore, nearly all parametric generalized index models have increasing link functions, making it a very natural assumption. In this talk, I will discuss estimation, computation, as well as rates of convergence under different underlying assumptions for the isotonic index model. This is joint work with Fadoua Balabdaoui from ETH Zurich and Cecile Durot from Paris X Nanterre.