Seasonal adjustment of weekly series is rarely done by statistical agencies. Few agencies publish weekly data - the Bureau of Labor Statistics publishes seasonally adjusted weekly UI claims.
In this paper, we'll examine the method BLS uses, and develop an R interface for the Fortran code. We will also examine a model-based method that develops the weekly seasonal adjustment from daily time series, and compares these adjustments to those generated by weekly numbers. Outliers and holiday effects will be included in this model. A similar model will be developed for weekly series where the daily observations are not available.
Several series from the US Federal Reserve and Statistics New Zealand will be used in the study.
|