Draft metadata profile
This draft implements a recommendation of the program: that metadata be a profile of four existing standards. It is a draft: no repository has been checked against it yet, and the program will revise it after the first rerun study. A JSON Schema (a machine-readable list of the fields each file must have) that checks it is not written yet.
Each standard does one job
| File | Standard | Fields a companion repository fills |
|---|---|---|
|
Citation File Format 1.2.0 |
|
|
CodeMeta 4.0 |
|
|
RO-Crate 1.3 |
The files of the repository, the data or its link, and the declared outputs below |
The DOI record |
DataCite Metadata Schema 4.7 |
Filled in by the archive, such as Zenodo, when the release is archived; |
The declared outputs are one RO-Crate action
None of the four standards says which file makes which table and figure. RO-Crate can say it with things it already has: an action (CreateAction) whose instrument is the entry point, whose object is the data, and whose results are the output files, each described with the table or figure of the paper it holds.
{
"@id": "#rerun",
"@type": "CreateAction",
"name": "Run the analysis from data to results",
"instrument": {"@id": "code/run_all.R"},
"object": [{"@id": "https://doi.org/10.xxxx/example-data"}],
"result": [{"@id": "outputs/table-2.csv"}, {"@id": "outputs/figure-3.png"}]
},
{
"@id": "outputs/table-2.csv",
"@type": "File",
"name": "Table 2",
"description": "Table 2 of the paper: time in range by treatment group"
}
The DOIs and file names above are placeholders, not a real study.
The profile can’t show the code is correct
It doesn’t say whether the code is correct or whether the paper’s conclusions follow from the results. It only lets a person or a program find the six parts and compare a rerun with the paper.