Estimation of seasonal factors for a short time span using multi-level modelling.
by
Craig McLaren
Australian Bureau of Statistics
Coauthors: Xichuan Zhang (Australian Bureau of Statistics)

Changes to the source or method of a survey can impact on the original survey estimates and the time series estimates, ie. trend and seasonally adjusted estimates, derived from them. To assess the impact of the change to the survey, a parallel survey original estimate can be calculated using data collected from the old and the new survey for one or more overlapping time periods. The number of overlapping time periods is typically short due to cost constraints. It is desirable to assist users by calculating time series estimates for the new survey. Traditional seasonal adjustment methods cannot adequately calculate time series estimates for short time series. Theory for estimating seasonal factors for short spans of time series data is given and illustrated with an example.

Date received: April 12, 2002