|
Activity Number:
|
116
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, August 3, 2009 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Statistics in Epidemiology
|
| Abstract - #304250 |
|
Title:
|
On Testing the Homogeneity of Across Quantile Changing-Point Functions
|
|
Author(s):
|
Nanshi Sha*+ and Ying Wei
|
|
Companies:
|
Columbia University and Columbia University
|
|
Address:
|
722 West 168th St. R-6, New York, NY, 10032,
|
|
Keywords:
|
quantile model ; change point ; cusum process
|
|
Abstract:
|
Piecewise linear quantile model with a changing point has been shown useful in many applications such as physiology and epidemiology. It is of natural interest to test whether the locations of change points are constant across the quantile levels. We focus on the case where two specific quantiles are of interest, i.e., whether the changing points at two given quantile levels $\tau_1$ and $\tau_2$ are constant. To test such hypothesis, we proposed to first estimate the quantile functions jointly assuming a common changing point, and then test the goodness-of-fit of the resulting quantile functions utilizing a cusum process of the residuals. The proposed test extends He and Zhu (2003), and is computationally efficient for changing-point quantile models. We studied its performance using Monte-Carlo simulations and applied to the real studies on adiposity rebound point as well as HIV study.
|
