Binds several scopus_records objects into a single one, renumbering
entry_number across the result and, optionally, dropping duplicates. This is
the safe way to merge separate fetches: plain rbind() would leave duplicate
entry numbers. c() on record sets calls scopus_combine(), so it renumbers
in the same way and, unless it is given dedupe, keeps every row.
Usage
scopus_combine(..., dedupe = FALSE)
# S3 method for class 'scopus_records'
c(x, ...)Arguments
- ...
One or more scopus_records objects, or a single list of them. A single set can be passed to de-duplicate it.
- dedupe
Whether, and by which rule, to keep once the records that are one record.
FALSE, the default, keeps every row.TRUE, also spelled"id_or_shared_doi", treats records as one when they share a 'Scopus' identifier, or when they share a DOI and one of them lacks an identifier."id_or_doi"also merges records with different identifiers that share a DOI, and"id_then_doi"applies the rule of the 0.4.0 release and earlier. See Details.- x
A scopus_records object (for the
c()method).
Value
A scopus_records tibble. Per-retrieval attributes such as
total_results and plan are not carried over, since they describe a
single fetch. The provenance a merge leaves true is kept, as
scopus_fetch_plan() keeps it for plan cells. When every input records
its retrieval time, retrieved_at is the earliest of the inputs' times and
retrieved_until the latest, so a baseline merged with its update is dated
from one to the other. Every contributing scopusflow_version is listed
when every input records one. Each is omitted when any input cannot supply
it, as a set read back from a .csv file cannot, since a value drawn from
the other inputs alone would misreport the set. The merge
itself is recorded in the combined attribute, a list of n_in
(records supplied), n_out (records kept), n_removed, deduplicated,
rule (the rule applied, NA when none was) and doi_conflicts (how
many DOIs records with different identifiers share, NA when no rule was
applied). It is what lets scopus_search_report() state how many
duplicates were removed and how, so the PRISMA-S item is answered.
Details
De-duplication compares DOIs as scopus_extract_dois() does, without
resolver prefixes such as https://doi.org/ or doi: and ignoring case. It
keeps the first record of each group, in the order the sets were given, and
a record with neither an identifier nor a DOI cannot be matched and is kept.
Under the default rule, records with the same 'Scopus' identifier are one
record, and a record without an identifier joins the identified record that
carries its DOI. Records with different identifiers are never merged, even
when they share a DOI. 'Scopus' holds duplicate records of one paper
(Valderrama-Zurián et al., 2015), but it also gives one DOI to distinct
papers (Franceschini et al., 2015), and a review loses a wrongly removed
record for good where a surviving duplicate is caught at screening. Such
DOIs are counted in doi_conflicts, which the search record reports, so
that they can be checked by hand. A record without an identifier whose DOI
two identified records carry is kept once, apart from both. Records without
an identifier that share a DOI no identified record carries are one record.
"id_or_doi" merges every record linked to another by a shared identifier
or a shared DOI, directly or through a chain of such links. It suits a set
whose shared DOIs have been checked and found to be duplicates.
"id_then_doi" keys each record on its identifier, failing that on its DOI
lower-cased but not otherwise cleaned, and so reproduces a set
de-duplicated by the 0.4.0 release or earlier.
References
Franceschini, F., Maisano, D., & Mastrogiacomo, L. (2015). Errors in DOI indexing by bibliometric databases. Scientometrics, 102(3), 2181–2186. doi:10.1007/s11192-014-1503-4
Valderrama-Zurián, J.-C., Aguilar-Moya, R., Melero-Fuentes, D., & Aleixandre-Benavent, R. (2015). A systematic analysis of duplicate records in Scopus. Journal of Informetrics, 9(3), 570–576. doi:10.1016/j.joi.2015.05.002
See also
scopus_fetch_plan(), which combines plan cells the same way.
Examples
# A baseline retrieval and a later one, merged into a cumulative set. The
# bundled corpus of real articles stands in for both, since 'Scopus'
# records may not be redistributed.
baseline <- example_records[example_records$year <= 2023, ]
later <- example_records
combined <- scopus_combine(baseline, later, dedupe = TRUE)
nrow(combined)
#> [1] 149
# Those records carry no 'Scopus' identifier, so de-duplication falls back
# to the DOI. The eleven that arrived without one cannot be matched, and so
# survive in both copies, which is why 138 distinct articles come back as
# 149 rows.
sum(is.na(example_records$doi))
#> [1] 11