Assessment of Uncertainty of Nonparametric Median Regression Estimators with Applications to Protein Lysate Arrays

• Activity Number: 68
• Type: Contributed
• Date/Time: Sunday, August 2, 2009 : 4:00 PM to 5:50 PM
• Sponsor: Section on Nonparametric Statistics
• Author(s): Xingdong Feng*+ and Xuming He and Jianhua Hu
• Companies: University of Illinois at Urbana-Champaign and University of Illinois at Urbana-Champaign and The University of Texas M.D. Anderson Cancer Center
• Keywords: Nonparametric ; EC50 ; Lysate array ; Wild bootstrap

Abstract

The protein lysate microarray is an emerging technology to directly measure protein concentration in biological samples. A nonparametric median regression model was proposed by Hu, et al. 2007 to quantify protein lysate arrays, but there has been no discussion on assessing the variability of the protein concentration estimates. We adopt the wild bootstrap method, which is capable of accounting for heteroscedasticity in the model, for variance estimation. However, the existing theory on the wild bootstrap applies only to linear estimators. In this paper, we propose a class of weight distributions that yield valid wild bootstrap variance estimates for median regressions. It is interesting to note that the weight distributions used in the classical wild bootstrap literature would lead to biased variance estimates for median regression.