Multivariate Quantile-on-Quantile Regression

Implements Multivariate Quantile-on-Quantile Regression (m-QQR) of Sinha, Ghosh, Hussain, Nguyen and Das (2023) , extending the bivariate Quantile-on-Quantile regression of Sim and Zhou (2015) to include exogenous moderators and controls with optional interaction terms. For each pair of quantile levels (theta of the response and tau of the regressor) the package fits a locally-weighted quantile regression of y on the principal regressor x, a lagged dependent variable, moderators Z and the x*Z interaction terms, using Gaussian kernel weights on the empirical cumulative distribution function (CDF) distance. Bootstrap standard errors and Koenker-Machado pseudo R-squared are reported. Visualisations include 'MATLAB'-style 'Parula' and 'Jet' 3D surfaces, heatmaps and contour plots through 'plotly'.