B0791
Title: Monotone single index models for conditional quantiles
Authors: Roger Koenker - University of Illinois (United States) [presenting]
Abstract: Transformation models assert that some linear function of observable covariates when transformed by a univariate function adequately approximates some functional of an observable random response, usually its conditional expectation. When the function is assumed to be monotone, such models can be estimated by profile likelihood; analogous methods for estimating conditional quantile functions will be described and rates of convergence for various target function classes will be discussed.