What are the basic concepts and terminology used in a time series repository and, specifically, Data Buffet?
This content originally appeared in the February 2012 issue of Data Buffet Monthly and has subsequently been updated.
The first concept is a time series database. Each time series consists of a sequence (or vector) of numeric observations (or periods) at a fixed frequency, possibly with gaps denoting holidays or other interruptions and represented with "ND" ("no data") values. Time series are amenable to operations (frequency conversion and analytic transformation) that would be onerous with a spreadsheet. For an indicator, a time series explains how much? and when?
The time series vector is clustered with a unique identifier (mnemonic) and explanatory metadata to form a time series object. Without these, the time series would be anonymous and undefined.
In Data Buffet, the metadata fields include a textual description (what?) and source citation (says who?) and timestamps for the start date and end date of the vector and when the object was last updated. The object has one canonical mnemonic and zero or more alias mnemonics (or aliases). Each mnemonic (pronounced: “new-mawn-ik”) consists of a concept code and geography code, which denote the what? and where? of the series. Some data vendors use numeric identifiers with no inherent meaning, but our mnemonics have structure, facilitating use of wild card expressions to retrieve multiple series related by topic or locale. Additionally, baskets (see below) can use prefab or custom geography lists to retrieve a specific set of locales.
Atop this repository we cluster series into releases, each of which is tied to long-form documentation (the how?), news articles and an advance release calendar. We use the catalog tree to organize concept codes by some combination of country, source, topic and more specific attributes, which may not otherwise be represented in the mnemonic.
The repository supplies all of our written and interactive products, including the DataBuffet.com web site. Data Buffet offers search modes for keywords, mnemonics, and find in catalog or you can browse the catalog tree, either the standard tree or custom filtered catalogs that you define and save. Series metadata and release documentation are consolidated in Mnemonic 411, whereas the reference files are standalone. To retrieve numeric data, assemble a basket; to visualize, plot a chart, pie chart, map or cycle comparison. All these workpieces offer frequency conversion, analytic transformation and arbitrary formulas. You can save workpieces for later, share with a group, and schedule them to run and deliver their output. Finally, by installing the Data Buffet Power Tools for Microsoft Office desktop software, you can pull live copies into your Excel, Word and PowerPoint documents.
Other interrelated subsystems manage client access through series subscriptions, the support tickets generated from LiveHelp! chats, and our routine data-harvest workflow. Our quality assurance system flags anomalies at all stages (series, metadata, release, calendar, documentation). You don't see these stages individually, but you benefit from timely updates, data consistency, and prompt support replies.