Nonparametric C- and D-vine-based quantile regression

Abstract Quantile regression is a field with steadily growing importance in statistical modeling. It complementary method to linear regression, since computing range of conditional quantile functions provides more accurate modeling the stochastic relationship among variables, especially tails. We introduce nonrestrictive and highly flexible nonparametric approach based on C- D-vine copulas. Vine copulas allow for separate marginal distributions dependence structure data can be expressed through graphical consisting sequence linked trees. This way, we obtain model that overcomes typical issues such as crossings or collinearity, need transformations interactions variables. Our incorporates two-step ahead ordering by maximizing log-likelihood tree sequence, while taking into account next two levels. show estimator consistent. The performance proposed methods evaluated both low- high-dimensional settings using simulated real-world data. results support superior prediction ability models.