Information model
SDMX in 2 minutes
To derive meaning from statistical data, understanding the associated
concepts is essential. For instance, the number 1.2953, on its own,
is meaningless. However, if we know it represents the exchange rate for
the US dollar against the euro on 23 November 2006, the data becomes
more meaningful.
Concepts are categorized into three types based on their roles:
Dimensions: Identify and describe the data. The combination of dimension values uniquely identifies statistical data (similar to a primary key). For exchange rates, examples of dimensions include numerator and denominator currencies, exchange rates types, and reference periods.
Attributes: Do not contribute to data identification but provide useful descriptive information. For example, the confidentiality status of statistical data.
Measures: Hold the values we care about, i.e. the statistical data.
Concepts usually have an expected type (representation). For instance, currencies are typically represented using ISO 3-letter currency codes. Many concepts expect a finite number of codes (codelist in SDMX), while others have characteristics such as data type, minimum and maximum length, or regular expressions (facets in SDMX).
All data in a statistical domain belong to a dataset (dataflow in SDMX). Data are reported by various providers and must comply with the expected structure of that dataflow (data structure definition or DSD in SDMX). The structure specifies relevant concepts, their roles, and expected types.
Metadata-driven processes
While the above outlines the metadata needed to describe statistical data, SDMX boasts a rich information model with various metadata types. These metadata can power diverse statistical processes, including data collection, validation, and mapping.
pysdmx provides an opinionated implementation of a subset of the SDMX
information model through Python classes. These classes enable the definition
of APIs for statistical data processes.
To efficiently support these processes, pysdmx model classes can be serialized in various formats, such as JSON, YAML, or MessagePack.
Refer to the pages below for detailed descriptions of the key model classes
defined by pysdmx.