Skip to contents

Attributes publication output (and citation impact) to entities using either full or fractional counting. Reviewers routinely ask for fractional counts so that a single multi-author / multi-institution paper is not counted in full for every contributor.

Usage

sm_count(
  corpus,
  method = c("full", "fractional"),
  level = c("institution", "author", "source"),
  call = rlang::caller_env()
)

Arguments

corpus

An sm_corpus.

method

"full" (default) gives each entity credit 1 per work it appears on. "fractional" splits each work's single unit of credit equally among the distinct entities on that work.

level

Entity level: "institution" (default), "author", or "source".

call

Caller environment for error reporting.

Value

A tibble, one row per entity, sorted by credit descending:

entity_id

Entity identifier.

entity_name

Human-readable name (falls back to entity_id).

n_works

Number of distinct works the entity appears on (this is the full count and is identical for both methods).

credit

Output credit: n_works under "full"; the sum of per-work fractional shares under "fractional".

weighted_citations

Citation impact attributed to the entity: summed cited_by_count under "full"; summed fractional-share weighted cited_by_count under "fractional".

Type-stable: an empty/inapplicable corpus returns a 0-row tibble with these columns.

Details

The fractional rule is the standard equal-share rule: a work with \(k\) distinct entities contributes \(1/k\) of credit (and \(1/k\) of its citations) to each. For level = "source" each work has a single source, so fractional and full counts coincide.

For level = "institution" each authorship row is labelled on its own, by the first of these authorships columns that has a value in that row: institution_match (from sm_affiliation_match()), institution_name, institution_id (named by corpus$institutions$display_name when that table lists the id) and raw_affiliation. A row labelled through institution_id has that id as entity_id; every other row has its label as both entity_id and entity_name. Rows labelled by their raw affiliation string are counted un-disambiguated, with a warning, because sm_count() applies no affiliation dictionary itself: run sm_affiliation_match() with a dictionary of your institutions first to count them. The order is the one sm_coverage_audit() uses for its affiliation levels, which however takes one column for all rows and falls back to country codes instead of raw strings. Up to scimapR 0.4.0, sm_count() used one column for every row (the first of institution_id, institution_match and raw_affiliation with any value), so rows without a value in it were left out, and it matched raw strings against the shipped sm_affiliation_dict before counting them.

sm_attribute_institution() keeps institution_match, so its roll-up of the matches to a controlled vocabulary (ROR ids and names in institution_id and institution_name) is counted only for rows without a match; up to scimapR 0.4.0 the counts followed institution_id. To count by the vocabulary, put its names first:

a <- corpus$authorships
corpus$authorships$institution_match <-
  dplyr::coalesce(a$institution_name, a$institution_match)

Examples

corpus <- sm_example_corpus(n_works = 30, seed = 1)
sm_count(corpus, method = "fractional", level = "author")
#> # A tibble: 58 × 5
#>    entity_id  entity_name      n_works credit weighted_citations
#>    <chr>      <chr>              <int>  <dbl>              <dbl>
#>  1 A000000001 Fatima Liu            30  7.04              100.  
#>  2 A000000002 Fatima Andersson       5  0.910               7.33
#>  3 A000000015 Sarah Tanaka           3  0.811               9.39
#>  4 A000000020 Maria Smith            3  0.792              15.0 
#>  5 A000000025 Erik Wang              3  0.783               4.58
#>  6 A000000039 James Patel            4  0.761              10.0 
#>  7 A000000016 Mohammed Brown         3  0.7                 7.2 
#>  8 A000000050 Mei Santos             3  0.65               14.4 
#>  9 A000000006 Fatima Brown           3  0.644               3.42
#> 10 A000000040 Lars Garcia            2  0.625               2.88
#> # ℹ 48 more rows

# Institutions: match the affiliation strings with a dictionary of your
# institutions first, then count
a <- corpus$authorships
corpus$authorships$raw_affiliation <- ifelse(
  a$position == 1L,
  "Unit North, Northtown",
  "Department of Surgery, Unit South"
)
dict <- list("Unit North" = "unit north", "Unit South" = "unit south")
matched <- sm_affiliation_match(corpus, patterns = dict)
#> ✔ Affiliation matching flagged 145 authorships across 2 institutions.
#> ℹ By signal: name_token: 145. See `sm_affiliation_summary()` for the full
#>   breakdown.
sm_count(matched, method = "fractional", level = "institution")
#> # A tibble: 2 × 5
#>   entity_id  entity_name n_works credit weighted_citations
#>   <chr>      <chr>         <int>  <dbl>              <dbl>
#> 1 Unit North Unit North       30     16               226.
#> 2 Unit South Unit South       28     14               208.