Beschreibung:
We provide a theoretical result for the time series-based cross validation that sheds light on the choice of validation sample size. We also consider an alternative way to construct validation samples and demonstrate the improved performance in certain situations via simulations