FUNCTION-ON-FUNCTION LINEAR QUANTILE REGRESSION

In this study, we propose a function-on-function linear quantile regression model that allows for more than one functional predictor to establish flexible and robust approach. The proposed is first transformed into finitedimensional space via the principal component analysis paradigm in estimation phase. It then approximated using estimated functions, parameter of constructed based on scores. addition, Bayesian information criterion determine optimum number truncation constants used decomposition. Moreover, stepwise forward procedure are significant predictors including model. We employ nonparametric bootstrap construct prediction intervals response functions. finite sample performance method evaluated several Monte Carlo experiments an empirical data example, results produced by compared with ones from existing models.