Efron-Petrosian integrals for doubly truncated data with covariates: an asymptotic analysis (Q2214246): Difference between revisions

Double truncation refers to situations, in which the target variable is subject to left and right random observation limits. Doubly truncated data require corrections for the observational bias, and this affects a variety of settings (including the estimation of marginal and multivariate distributions, regression problems, and multi-state models). The paper under review addresses theoretical developments for the random double truncation model. The authors review relevant previous publications and present their new results. Namely, Efron-Petrosian integrals for doubly truncated data with covariates are introduced. These are empirical multivariate integrals with respect to the marginal nonparametric maximum likelihood estimator of the doubly truncated variable. The authors' main derived result is the consistency of the Efron-Petrosian integrals, together with an asymptotic i.i.d. representation, which entails convergence to a normal distribution. Application of the main result to the asymptotic analysis of the Pearson correlation coefficient, least squares regression, Cox regression, and cumulative incidence functions is also given. The finite sample performance of the Efron-Petrosian integrals is investigated through simulations.