As most of you know, you can generate estimation and prediction study packages directly from ATLAS. However, the current version of ATLAS is lagging behind a bit in the versions of the HADES packages used in the generated study packages. Most importantly, the study packages generated by ATLAS still use ff
, even though we migrated to Andromeda
for good reasons a while ago.
If you want to use the latest versions of the study packages (which use the latest version of the HADES packages), you can generate the study packages from the ATLAS JSON yourself. This is done through the Hydra package. See these instructions for how to use Hydra to generate study packages based on the ATLAS JSON.
Not only do the new versions use Andromeda, avoiding issues with ff like this one and many others related to zero-row tables, they also use the latest version of Circe to generate the cohort SQL (thus supporting BigQuery). The estimation study package now also uses renv to handle package dependencies (see the README file in the generated study package).