|Page tools: Print Page Print All RSS Search this Product|
Future seasonal adjustment infrastructure
Keeping seasonal adjustment infrastructure efficient, methodologically modern and quality assured is always a focus for the Time Series Analysis section (TSA), but more so in light of the current emphasis on building systems that can be used across many statistical production areas of the ABS, handle future metadata requirements and easily interact with other systems. An essential part of any roadmap to new seasonal adjustment infrastructure is ensuring that future inputs, seasonal adjustment 'engines' and outputs are handled efficiently and are adaptable. Further, evaluations of what existing seasonal adjustment infrastructure still has to offer versus other alternatives, be they pure 'off the shelf products' or 'best of both world cross breeds', need to be conducted. This necessitates a more flexible 'sand pit' environment for external software feasibility studies. A series of Methodological News articles over coming issues will outline key aspects of the future seasonal adjustment infrastructure plan.
This article will focus briefly on current time series metadata, referred to as 'series knowledge'. Series knowledge contains, besides the obvious start/end date and periodicity variables, a detailed record of analysis parameters required for appropriate seasonal adjustment and trending, aggregation structures to achieve seasonally adjusted and trend Australian level series from their respective states and industries (for example) and uniquely, existing software SEASABS has the ability to store past analyses in the form of archives.
Exploring such metadata, determining what is no longer relevant, determining what will be needed in the future and successfully translating it from text file format to an international standard like SDMX/DDI are all important first steps in ensuring greater compatibility between seasonal adjustment infrastructure and other systems. Standard time series metadata allows seamless connectability between systems be they processing based (e.g. FAME or SAS) or analytical tools (e.g. Demetra, Win X13 or R) and this is a key aspect of streamlining future time series processes. New metadata is envisaged to include greater representation of related time series across collections, improved quality assessment parameters, associated bonafide quality gate flags (which can be also monitored over time) and parameters from new knowledge / business rules to successfully govern envisaged batch capability (seasonal adjustment of a group of time series).
A series knowledge 'translator' (from existing text file format to SDMX/DDI) is expected to evolve as part of the move towards a standard sand pit evaluation environment through current collaboration between TSA and the ABS data management and technical application areas. Such a sandpit is to facilitate evaluation of not just time series software, but also other software that other parts of the Division wish to bring in to the ABS environment, aiming to significantly streamline this process within the ABS.
These documents will be presented in a new window.