1.0.77
UDP Marts
Last updated
UDP Marts
Last updated
Release Date
October 30, 2024
A new datamart exists in every prod tenant's BQ environment called . This mart defines the categorical and continuous fields in the UDP context store. The purpose of this mart is to characterize the types and distribution of values present in the UDP in aggregate. Also, these distributions are the basis for the creation of the synthetic data. Introducing these marts in each tenant's prod environment increases the transparency of how the synthetic data is generated.
There are four tables in the udp_distributions dataset in BQ:
categorical_distributions - fields in the UCDM with discrete, countable values
continouus_distributions - fields in the UCDM with continuous, potentially unbounded values.
categorical_dependent_distributions - categorical fields with dependency on at least one other UCDM field
context_store_distributions - a union of the first three tables
For categorical fields, we focus on counting num_records and pct_records to determine the relative proportion of each countable value. For continuous fields, average, median, and standard deviation are calculated.
These tables are not necessarily complete, and as we get feedback, we will add fields and improve them. The goal for these marts are to have an accurate summary view of the data in each tenant's UDP. More details on these marts will be shared at the Unizin Townhall on November 12, 2024.
The marts previously had a filter that limited the computation to only undergraduate students. This has been removed, and now all types of student enrollment can be supported by the student activity score.
The fields academic_level_code and academic_level_description are added to the schemas of the tables to identify the type of enrollment.