Data-driven P-Splines under Short-Range Dependence

This paper focuses on data-driven selection of the smoothing parameter in P-splines for time series with short-range dependence. Well-known asymptotic results of P-splines are first adapted to the current context. A fully automatic iterative plug-in (IPI) algorithm for P-splines is investigated in a comprehensive simulation study. Practical relevance of the IPI is shown by application to economic time series. Moreover, it is illustrated that the IPI can be used for automatic selection of the smoothing parameter of the Hodrick-Prescott filter. Furthermore, a P-spline Log-ACD model is proposed and applied to average daily trade duration data. Smoothing parameter selection is carried via the proposed IPI-algorithm, which performs very well in this context too