Statistical inference for the functional quadratic quantile regression model (Q2202044)

To investigate a linear relationship between the mean or quantile of a scalar response and a functional predictor defined on a compact set of \(\mathbb R\), functional linear or quantile regression models have been proposed. \textit{F. Yao} and \textit{H.-G. Müller} [Biometrika 97, No. 1, 49--64 (2010; Zbl 1183.62113)] extended the functional linear model by adding a quadratic term of the functional predictor. A similar extension to the functional quantile regression model is done by the authors in this paper. A procedure based on functional principle components is proposed to estimate the intercept and functional coefficients in the model. Asymptotic normality of the estimated intercept and convergence rates of the estimated functional coefficients are established. A rank score procedure is also proposed to test whether the coefficient for the quadratic functional term is significantly different from 0. The proposed procedures are evaluated through simulation studies and also applied to study the relationship between the fat content and absorbance curve observed on a sample of meat.