New method improves accuracy of time series models for economic forecasting.
The article explores different methods to estimate bandwidths in time series models. They use a Bayesian approach to select bandwidths for local linear estimation, which outperforms traditional methods. The researchers also develop a model with trending regressors and propose a Bayesian method for bandwidth selection, showing better results than cross-validation. Lastly, they introduce a new bandwidth estimator for kernel density estimation based on mixing processes.
