CMStatistics 2017: Start Registration
View Submission - CMStatistics
B0483
Title: Testing in heteroscedastic quantile varying coefficient models Authors:  Irene Gijbels - Katholieke Universiteit Leuven (Belgium)
Mohammed Abdulkerim Ibrahim - Hasselt University (Belgium)
Anneleen Verhasselt - Hasselt University (Belgium) [presenting]
Abstract: Quantile regression in varying coefficient models is considered. Varying coefficient models are a flexible generalization of linear regresion models, in the sense that the coefficients are functions (of a covariate) instead of constants. Varying coefficient models have been used for modeling longitudinal data, where the coefficients are then allowed to vary with time. The coefficient functions are estimated with splines. We focus on developing a testing procedure for the shape (constant, monotone, convex) of the coefficient functions. Furthermore, we allow the error structure to be heteroscedastic and propose a likelihood-ratio-type testing procedure for comparing the variability functions. The performance of the procedures is investigated via simulation studies and their use is illustrated on data examples.