Local Composite Quantile Regression Smoothing: Flexible Data Structure and Cross-validation
Xiao Huang and Zhongjian Lin
Coles Working Paper Series, SPRING18-04, March 2018
Overview This paper examines local polynomial regression based on averaging different quantile estimates for both continuous and categorical variables, such as gender and race. Local polynomial regression is widely used to model nonlinear patterns in the data. The proposed estimator can improve estimation efficiency by averaging different quantile estimates. It is applicable to cases whenever we need to estimate a regression function and the data exhibit nonlinearity. We also propose a data-driven method to select bandwidths to further improve the estimator.
32 | Working Papers
Made with FlippingBook - Online catalogs