Enhanced flagging of imputed child data records

The CUSP data model includes both real and imputed (or synthetic) data. Which is why, for the purpose of transparency, auditability, and reliability of our data and analyses, we clearly flag data records that are imputed, even for children for whom actual data exists in state data systems. The most recent release of the CUSP SDK includes enhancements in our imputed data flagging capability to empower our clients to use and interpret data as correctly and knowledgeably as possible.

FEATURES
  • New table in the CUSP output schema clearly and comprehensively identifies which data records—address, age, household income, or household employment status—are imputed for children present in state administrative data systems.
  • The table retains original values of data records, imputed values, and final selected values for every attribute of children from administrative data systems.

Benefits

  • Greater transparency and auditability of child-level data, with full visibility into synthetic records for real children.
  • Retaining original records (including null and faulty records) alongside imputed records and clearly displaying which value the model used—original or imputed—provides full transparency on the decision rules employed in the model and the types of data quality issues that they address with imputed records.