Building on a COVID-19 modeling project being undertaken by the DELPHI epidemiological forecasting team at Carnegie Mellon University, Daniel McDonald and his team will develop and implement statistical procedures that flag potential data quality issues before it is used by DELPHI or released to the public through the API. A challenge the modelling team at DELPHI faces is the rapid identification of anomalous indicator values, which can cause data quality issues for internal and external data consumers since this data are made public without any lag. This makes it difficult to systematically and rapidly catch these issues.
The results of this project will be used internally by the modelling team and integrated into the forecast pipeline. Additionally, his team will work with the visualization team to incorporate the developed diagnostic tools into the Covidcast site for the benefit of external data consumers.