On Rank Score Test for Longitudinal Bent Line Quantile Model

• Activity Number: 43
• Type: Contributed
• Date/Time: Sunday, August 1, 2010 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Nanshi Sha*+
• Companies: Columbia University
• Keywords: Quantile regression ; Longitudinal ; Rank score test ; change-point

Abstract

Bent line quantile model has been shown useful in many applications such as physiology and epidemiology. In this article we consider developing a robust rank score test for longitudinal bent line change-point quantile regression model. We propose two reliable tests for detecting location of change-point and slope coefficient. Through a series of simulation studies we demonstrate that the confidence intervals generated by inverting the proposed rank score tests have certain advantages over those obtained from bootstrap method in terms of shorter lengths, less computational burden and highly robustness against heteroscadesticity. We illustrate the use of the proposed tests by applying them to a real longitudinal HIV study. Estimates and confidence intervals of the treatment-plateau-point are obtained and the long-term treatment effect are evaluated.