|
Activity Number:
|
276
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 4, 2009 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Biometrics Section
|
| Abstract - #305668 |
|
Title:
|
Quantile Regression for Longitudinal Biomarker Data Subject to Detection Limits
|
|
Author(s):
|
Lan Kong and Minjae Lee*+
|
|
Companies:
|
University of Pittsburgh and University of Pittsburgh
|
|
Address:
|
Dept of Biostatistics, Pittsburgh, PA, 15261,
|
|
Keywords:
|
quantile regression ; biomarker ; detection limit ; longitudinal data ; left-censored
|
|
Abstract:
|
Biomarkers are often measured repeatedly in biomedical studies to better understand the natural history and development of the disease. One common source of measurement error for biomarkers is left-censoring due to the detection limits. Usually multivariate normal distribution was assumed to account for censored observations. However, the biomarker data are often highly skewed even after certain transformations. The quantile regression provides a useful alternative by imposing minimal assumption on the distribution and also allows one to relate the change of different quantiles to the covariates differently. We develop the estimating procedures for left-censored longitudinal data based on a marginal quantile regression model and provide a bootstrap method for the variance estimation. We evaluate the proposed methods through simulation studies and use a sepsis data set for demonstration.
|
