The ref_summary module does not run a check of its own. Instead it collects the results of the other reference-section modules — accuracy, PubPeer, replication, and retraction — into a single table with one row per reference, so you can see everything known about each cited work in one place.
Because it summarises the output of those modules, ref_summary must run after them, in the same report. On its own it has nothing to summarise.
The modules it summarises make live network calls, so an internet connection is needed.
The summary needs the other reference modules to have run first. The natural way to arrange that is to run them together with report_module_run(), which keeps each module’s output and passes the chain along (see Creating a Metacheck Report). List ref_summary last:
paper <- demopaper()
module_output <- report_module_run(
paper,
modules = c("ref_accuracy", "ref_pubpeer",
"ref_replication", "ref_retraction",
"ref_summary")
)
mo <- module_output[["ref_summary"]]
mo$traffic_light#> [1] "info"mo$summary_text#> [1] "Summary information provided for 5 references"The combined table has one row per reference and one column per upstream check:
| bib_id | accuracy_mismatch | pubpeer | replication_type | retractionwatch |
|---|---|---|---|---|
| 0 | no match | NA | NA | NA |
| 1 | NA | NA | replication | NA |
| 2 | title | Link | NA | Retraction |
| 3 | doi, author | NA | NA | NA |
| 4 | no match | NA | NA | NA |
A reference is interesting when any of these columns is filled in: a reported accuracy mismatch, a PubPeer comment, a replication outcome, or a Retraction Watch entry. Follow up those references using the dedicated chapter for whichever column flagged it.
ref_summary inherits everything from the modules it summarises, so the same cautions apply: a filled-in cell is a prompt to look, not a verdict. Its value is convenience — instead of reading four separate tables, you scan one. The summary itself is informational (info); the underlying modules carry the traffic lights.
ref_summary takes only the paper argument, but it only produces useful output when the other ref_* modules have run before it in the same report.