Midastar: Threshold Autoregression with Data Sampled at Mixed Frequencies

We propose Midastar models by combining the Mixed Data Sampling (MIDAS) and the threshold autoregression (TAR). The Midastar model of the first kind is designed for a low frequency target variable and a high frequency threshold variable. The proposed model can detect threshold effects accurately, while the aggregated TAR has a risk of finding spurious non-threshold effects. The Midastar model of the first kind has desired asymptotic and finite sample properties. We apply the proposed model to Japan's COVID-19 data, detecting significant threshold effects. We also propose and elaborate the Midastar model of the second kind designed for a high frequency target variable and a low frequency threshold variable.