Probabilistic index models

Authors

Olivier Thas, University of WollongongFollow
Jan De Neve, Ghent University, Belgium
Lieven Clement, Ghent University, Belgium
Jean-Pierre Ottoy, Ghent University, Belgium

RIS ID

64781

Publication Details

Thas, O., De Neve, J., Clement, L. & Ottoy, J. (2012). Probabilistic index models. Journal of the Royal Statistical Society Series B: Statistical Methodology, 74 (4), 623-671.

Abstract

We present a semiparametric statistical model for the probabilistic index which can be defined as P (Y <-Y*) where Y and Y* are independent random response variables associated with covariate patterns X and X* respectively. A link function defines the relationship between the probabilistic index and a linear predictor. Asymptotic normality of the estimators and consistency of the covariance matrix estimator are established through semiparametric theory. The model is illustrated with several examples, and the estimation theory is validated in a simulation study.