Regression analysis of doubly truncated data

TL;DR

This paper introduces a new regression estimation method for doubly truncated data, extending rank-based estimators, with proven consistency, asymptotic normality, and demonstrated effectiveness through simulations and real data analysis.

Contribution

It proposes a general, easily implementable regression estimator for doubly truncated data, extending existing rank-based methods and providing theoretical guarantees.

Findings

Estimator is consistent and asymptotically normal.

Method performs well in simulations.

Re-analysis of astronomical data demonstrates practical utility.

Abstract

Doubly truncated data are found in astronomy, econometrics and survival analysis literature. They arise when each observation is confined to an interval, i.e., only those which fall within their respective intervals are observed along with the intervals. Unlike the more widely studied one-sided truncation that can be handled effectively by the counting process-based approach, doubly truncated data are much more difficult to handle. In their analysis of an astronomical data set, Efron and Petrosian (1999) proposed some nonparametric methods, including a generalization of Kendall's tau test, for doubly truncated data. Motivated by their approach, as well as by the work of Bhattacharya et al. (1983) for right truncated data, we proposed a general method for estimating the regression parameter when the dependent variable is subject to the double truncation. It extends the Mann-Whitney-type rank estimator and can be computed easily by existing software packages. We show that the resulting estimator is consistent and asymptotically normal. A resampling scheme is proposed with large sample justification for approximating the limiting distribution. The quasar data in Efron and Petrosian (1999) are re-analyzed by the new method. Simulation results show that the proposed method works well. Extension to weighted rank estimation are also given.