matthewhorridge
commented
4 months ago For starters:
Overall, make the metadata more specific to RADx. Having a general Datacite-like template is suboptimal for instance creation and consumption. Datacite-like instances/template can be generated programmatically.
Metadata cleanup and extension
- Add metadata fields for file metadata (file size)
- Add metadata fields for CSV descriptions (rows/records, fields/columns)
- Add fields to describe the list of variables (with links to a Global Code Book Ontology) present in a data file
- Add metadata field for derived from file (for transformcopy derived from origcopy)
- Add fields for descriptive statistics
- Add fields to specify whether the data is harmonized or not
- Add file to point to readme file (if present)
- Improve the mechanism (keywords and MeSH terms) for specifying data file contents
- Encode file creation dates directly
- Encode pointer to study metadata file (needs discussion whether to adopt this)
- Flatten out elements with single fields
- Add lists of conditions (ontology terms) for which there is data
Candidates for pruning:
Prune metadata fields for which we will never be able to collect any metadata
- Data streams element (never been filled out and too hard to do posthoc)
- Data file creation processes
- Data file temporal coverage
- Data file spatial coverage
- Data file elevation coverage
Auxiliary metadata
This feels pretty ugly at the moment. Needs review.
We have been working with the RADx metadata template and with DCCs for some time now and we could probably provide a more domain centered metadata template for the kinds of metadata that DCCs can actually provide. Furthermore, we can insert structured descriptive statistics for variables/DEs in data files. More details to follow.