Quantile regression-based seasonal adjustment

Time series of different nature might be characterised by the presence deterministic and/or stochastic seasonal patterns. By seasonality, we refer to periodic fluctuations affecting not only mean but also shape, dispersion and in general density variable interest over time. Using traditional approaches for adjustment efficient because they do ensure, instance, that adjusted data are free from behaviours in, say, higher-order moments. We introduce a method based on quantile regression is capable capturing forms Given interest, describing its behaviour an approximation entire conditional distribution, removing patterns variance, or varying quantiles distribution. In first part this work, provide proposed approach deal with pattern cases. empirical examples simulated real where compare our proposal least-squares approaches. The results favour case if change across quantiles. second improve flexibly account essential effect structural breaks time series. Again, methods segmented-least squares several affected both results, behaviour, especially when