|
Abstract:
|
The Covid-19 crisis has increased the interest in high-frequency series. It was indeed necessary to monitor as quickly as possible the effects of the various measures against the pandemic. Besides the usual difficulties linked to the treatment of high-frequency series, the crisis has generated strong breaks, which cannot always be easily tackled. Nonetheless, it appears that the usual seasonal adjustment algorithms can be adapted to extract from the series the relevant movements. We focus in this work on model-based approaches, like the canonical decomposition of specialized ARIMA models and specific structural models. Beside the description of the chosen models, we explain the solutions – mainly based on state space algorithms – used to solve the various technical challenges. The proposed methods are implemented in a R-package, freely available, and are applied on actual series.
|
