Resumen de Predicting using quantile regression methods
Mercedes Conde Amboage, Wenceslao González Manteiga 
, César Andrés Sánchez Sellero
Quantile regression methods are considered for computing predictions and prediction intervals of NOx concentration measured in the surroundings of the power plant of As Pontes (Spain). For these data, smaller prediction errors were obtained by median regression compared to mean regression. A new method to construct prediction intervals based on median regression and bootstrapping the prediction error is proposed. This new method rendered better coverage results for NOx data, compared to classical and bootstrap prediction intervals for mean regression, and also compared to simpler prediction intervals based on quantile regression. A simulation study is given to illustrate those features that, similarly to these particular data of NOx concentration, can be found in many other environmental datasets, and will lead to a better performance of the proposed method for obtaining prediction intervals.
