Forecasting in nonlinear univariate time series using penalized splines (Q1685198): Difference between revisions

The authors consider fitting and forecasting for three special models of the centered, strictly stationary process of order \(d\) with an autocorrelated structure given by \(Y_t=f(Y_{t-1},\dots,Y_{t-d})+\varepsilon_t\), where \(f\) is an unknown smooth function and \(\varepsilon_t\) is a white noise. In the first part of the manuscript, the authors apply the penalized spline smoothing to estimate the functional effects of the additive autoregressive model, the functional autoregressive model and the single index autoregressive model. The second part of the manuscript deals with forecasting of these time series models and finishes with nonparametric modeling of Euro overnight rate dynamics.