Statistics and Its Interface

Volume 2 (2009)

Number 1

Nonparametric methods for employment termination times with competing causes

Pages: 37 – 44

DOI: https://dx.doi.org/10.4310/SII.2009.v2.n1.a4

Authors

Werner A. Stahel (Seminar für Statistik, Swiss Federal Institute of Technology, Zürich, Switzerland)

Mara Tableman (Department of Mathematics & Statistics, Portland State University, Portland, Oregon, U.S.A.)

Abstract

A dataset of termination times of a company is analyzed by fitting a competing risk model, taking left truncation and right censoring into account. The analysis intends to provide evidence for a claim of age discrimination. The model imposes no functional form on the parameters of interest, namely the survivor, cause-specific hazard, and cumulative incidence functions. The empirical estimators are modified to account for delayed entry. Inference is primarily based on smoothed empirical hazard functions. The approach provides a frequentist alternative to the Bayesian analysis advocated by Kadane and Woodworth (2004) and is easy to implement using existing software.

Keywords

age discrimination, competing risks model, conditional probability function, flow data, kernel estimator of hazard, left truncation, nonparametric, time-varying covariate

2010 Mathematics Subject Classification

Primary 62N02. Secondary 62G05.

Published 1 January 2009