TVCAR models for forecasting

This paper predicts the 100 missing values in the CATS Benchmark using a Time-Varying Coefficient Autoregressive Model (TVCAR). The TVCAR model is an autoregressive model in which the coefficients vary smoothly with time. The model is fitted to the first differences of the data by minimising the residual sum of squared, subject to certain restrictions that enable the gaps left by the missing observations to be bridged. The path of each time-varying coefficient is initially described by a combination of cosine functions. Later, the method is improved replacing the cosine specifications by piecewise polynomials.