RESEARCH ARTICLE

Subdivisions and crossroads: Identifying hidden
community structures in a data archive’s
citation network

开放访问

杂志

Sara Lafia1

, Lizhou Fan2

, Andrea Thomer2

, and Libby Hemphill1,2

1ICPSR, 密歇根大学, 安娜堡, MI
2School of Information, 密歇根大学, 安娜堡, MI

引文: Lafia, S。, Fan, L。, Thomer, A。, &
Hemphill, L. (2022). Subdivisions and
crossroads: Identifying hidden
community structures in a data
archive’s citation network. Quantitative
Science Studies, 3(3), 694–714.
https://doi.org/10.1162/qss_a_00209

DOI:
https://doi.org/10.1162/qss_a_00209

已收到: 16 可能 2022
公认: 22 六月 2022

通讯作者:
Sara Lafia
slafia@umich.edu

处理编辑器:
Ludo Waltman

关键词: archival science, community detection, data citation, data reuse, network analysis

抽象的

Data archives are an important source of high-quality data in many fields, making them ideal
sites to study data reuse. By studying data reuse through citation networks, we are able to learn
how hidden research communities—those that use the same scientific data sets—are
organized. This paper analyzes the community structure of an authoritative network of data
sets cited in academic publications, which have been collected by a large, social science data
archive: the Interuniversity Consortium for Political and Social Research (ICPSR). Through
network analysis, we identified communities of social science data sets and fields of research
connected through shared data use. We argue that communities of exclusive data reuse form
“subdivisions” that contain valuable disciplinary resources, while data sets at a “crossroads”
broadly connect research communities. Our research reveals the hidden structure of data
reuse and demonstrates how interdisciplinary research communities organize around data sets
as shared scientific inputs. These findings contribute new ways of describing scientific
communities to understand the impacts of research data reuse.

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
q
s
s
/
A
r
t
我
C
e
–
p
d

我

F
/

3
3
6
9
4
2
0
5
7
7
8
5
q
s
s
_
A
_
0
0
2
0
9
p
d

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

介绍

1.
Data are essential resources for social science research, and data creators’ contributions
should be rewarded (Alter & 冈萨雷斯, 2018). In addition to ensuring credit, measures of data
reuse such as downloads and citations can reveal a data set’s role in a research community
and provide insights into how researchers engage with data (Cousijn, Feeney et al., 2019).
Analyzing data citations reveals data citation practices and provides a way to quantify the
analytical utility and disciplinary reach of data collections (Buneman, Dosso et al., 2021).
然而, it has typically been challenging to find these measures because download data
is not widely available, and researchers inconsistently cite data (Buneman, Christie et al.,
2020; Lowenberg, Chodacki et al., 2019). Incomplete or opaque research data citations fail
to include persistent identifiers, which create obstacles to tracking data use and fail to give
appropriate credit to data creators (Moss & Lyle, 2018).

Data archives—particularly domain-specific archives with robust curation services—are
ideal sites to study data reuse. They provide data services that make reuse easier, making them
sites of research convergence. Archives anticipate data sets that have high analytical potential
for long-term preservation and community impact as “topical collections” (Fenlon, 2017;
帕尔默, 韦伯, & Cragin, 2011). 此外, some maintain bibliographies of papers that

版权: © 2022 Sara Lafia, Lizhou
Fan, Andrea Thomer, and Libby
Hemphill. Published under a Creative
Commons Attribution 4.0 国际的
(抄送 4.0) 执照.

麻省理工学院出版社

Subdivisions and crossroads

reuse data from the archive, therefore tracking “citations” even when they are not formally
included in a paper (例如, NASA’s Data Archive Centers: DAACs1; biodiversity data aggrega-
tors such as Global Biodiversity Information Facility: GBIF2; and Data Observation Network
for Earth: DataONE3). There has been relatively little analysis of the intercitation networks
resulting from research data reuse in academic literature, 然而.

Citations of data in these archives create networks of data sets with attributes that help us
understand data reuse and its implications. 例如, understanding the context of data
discovery and reuse may help us understand the distribution of ideas or topics within and
between research domains, and identify data sets that exhibit exceptional long-term analytical
潜在的 (Palmer et al., 2011). Like “hibernators” among research papers (胡 & Rousseau,
2019), valuable data sets may lay dormant for years until they are discovered and “awakened”
through reuse. Identifying the different functions that data serve within knowledge communi-
ties can help us ensure data creators receive appropriate credit for their contributions.

此外, looking for new patterns of data couse or reuse would help identify hidden
communities that use archived data in novel ways. Data reuse can be viewed as an indirect
form of cooperation and collaboration between researchers (Sands, Borgman et al., 2012;
Thomer, Twidale, & Yoder, 2018; Zimmerman, 2008). Data archives promote research by pro-
viding access to data sets, and some of these data sets function as “boundary objects” (Star &
Griesemer, 1989) or parts of shared information spaces (Bannon & 施密特, 1989). The visi-
bility of data reuse depends on the vantage point; while data reuse may be visible to those
directly involved, larger patterns of reuse may remain invisible, both to the data archive
(例如, data managers) and to prospective data users from different disciplines. Revealing hidden
reuse communities and their structures helps us understand what roles data play in knowledge
production and how they function as boundary objects between fields of research.

Despite recent data-sharing mandates, securing data deposits is still a challenge for data
archives. Researchers are often wary of sharing data because they fear being “scooped” or
are unsure how other researchers might use their data (Borgman, Scharnhorst, & Golshan,
2018; Cragin, Palmer et al., 2010). Mapping the network of data citations provides evidence
of data reuse that will help data producers and archives better assess the collaborative utility of
data and demonstrate different types of secondary use to researchers and potential depositors.

在本文中, we inspect an authoritative bibliography of social science data sets cited in
academic publications from the Inter-University Consortium for Political and Social Research
(ICPSR) Bibliography of Data-related Literature4. 具体来说, we analyzed its citation graph to
uncover hidden community structures and identified the different roles data sets play in net-
worked communities. By linking citations to metadata from a scholarly database, Dimensions,
we were able to include attributes such as “fields of research”5 in our analysis (Hook, Porter, &
Herzog, 2018). We then used community detection algorithms to identify hidden communities
within the network of data citations and identified two types of data sets that unite scientists
involved in social science knowledge production: subdivision data sets and crossroads data
套. Subdivisions exclusively function as disciplinary resources used by a narrow set of fields.

1 https://lpdaac.usgs.gov/resources/publications/.
2 https://www.gbif.org/resource/search?contentType=literature&relevance=GBIF_USED.
3 https://search.dataone.org/profile.
4 https://www.icpsr.umich.edu/web/pages/ ICPSR/citations/.
5 According to Dimensions, the fields of research (FoR) is a hierarchical classification applicable for catego-

rizing all research and development activity.

Quantitative Science Studies

695

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
q
s
s
/
A
r
t
我
C
e
–
p
d

我

F
/

3
3
6
9
4
2
0
5
7
7
8
5
q
s
s
_
A
_
0
0
2
0
9
p
d

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

Subdivisions and crossroads

Crossroads, 相比之下, enable interdisciplinary research. The network structures we identify
and name acknowledge the variation in reuse and help us recognize the myriad functions that
data sets serve in scientific communities.

2. BACKGROUND AND RELATED WORK

2.1. Data Archives as a Site for Understanding Scholarly Communication Practices

Data archives support data-intensive research by providing long-term data stewardship,
使用权, and high-quality data curation. Notable examples of data archives with high levels
of curation include GenBank, a rich repository of genetic sequence data; SESAR, a repository
of metadata describing physical samples in the earth sciences, as well as links to derived data
套; and PANGAEA, a publisher for georeferenced data sets linked to earth system studies.
Data sharing through archives enables researchers to find and reuse data that they did not
collect. 换句话说, data created for one purpose can be used by new audiences to answer
new questions (棕色的, 2003; Wilkinson, Dumontier et al., 2016). Researchers can use existing
data to validate previous findings, extend their data collections, or form the basis for new stud-
ies via integration or independent reuse (Gregory, Groth et al., 2020; 国王, 1995; Pasquetto,
Randles, & Borgman, 2017; Thomer, 2022). 此外, as more funders and journals man-
date that data from grants and papers be shared openly, data archives are only growing in
importance as sites of scholarly communication.

The data held in these repositories often have untapped reuse potential across disciplinary
边界 (嘿, Tansley, & Tolle, 2009; Palmer et al., 2011). Such interdisciplinary research
using archived data can lead to breakthrough discoveries (美国国家科学院,
2005; Tenopir, Allard et al., 2011). Fields of research may share an interest in explaining dif-
ferent aspects of the same phenomenon, giving rise to interfield theories that bridge fields of
科学 (Darden & Maull, 1977). “Borderland disciplines” sometimes form where fields of
research collide over shared resources, such as instruments or data, leading to the evolution
of new techniques (Gökalp, 1987). Data sets that facilitate interactions between research areas
therefore function as “boundary objects,” carrying multivalent analytical potential across
research communities (Star & Griesemer, 1989) and facilitating knowledge exchange across
边界. 然而, there has been little research on the prevalence of such data sets-as-
boundary-objects. We know little about which features of data sets promote boundary crossing,
or how to measure their collaborative potential.

2.2. Data Citation Standards and Emerging Data Citation Networks

One way of exploring interdisciplinary data reuse—and therefore, the extent to which data sets
function as boundary objects between communities—is by studying data citation networks.
Efforts to promote data citation over the last 20 years have led to the adoption of new data
citation practices in many communities. Milestones formalizing data citation include the Joint
Declaration of Data Citation Principles (Data Citation Synthesis Group, 2014), Data Citation
Roadmap for Scholarly Data Repositories (Fenner, Crosas et al., 2016), and Data Citation
Roadmap for Scientific Publishers (Cousijn, Kenall et al., 2018). Data citation counts provide
a foundation for studying the scholarly impact of scientific data and the value of data curation
努力.

The adoption of data citation principles makes it possible to analyze emerging data reuse
behavior and structures of hidden research communities in data citation networks. 引文
networks generally represent documents as vertices and citations of one document by another

Quantitative Science Studies

696

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
q
s
s
/
A
r
t
我
C
e
–
p
d

我

F
/

3
3
6
9
4
2
0
5
7
7
8
5
q
s
s
_
A
_
0
0
2
0
9
p
d

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

Subdivisions and crossroads

as edges (光, Clarkson et al., 2007). Citation networks can highlight central nodes such as
influential institutions; heavy edges between nodes indicate important connections and pro-
过程, such as the diffusion of ideas (陈, 2017). Prior studies of citation networks have pro-
vided insights into ties between individual researchers and collaborations between research
学科 (Tomasello, Vaccario, & Schweitzer, 2017). Studies of publication citation networks
(例如, papers or journals) have also identified novel papers, measured the impact of papers and
their authors, and attributed discoveries to authors (纽曼, 2004).

Whereas publication citations broadly enable lineage retrieval for ideas, data citations indi-
cate the origins and processing history of the data sets that have been used in an analysis (Bose
& Frew, 2005). Data citation networks reflect connections between disciplinary literature and
the research data that they draw from. They reveal the reach of research data and support the
computation of bibliometrics that show the relationships and impacts of scientific products
(Buneman et al., 2021). The interactional context of data production and citation also reflects
relationships between data producers and consumers in a broader data economy (Vertesi &
Dourish, 2011).

Quantifying the scholarly impact of data archives and other research infrastructures relies on
proxy measures for data usage, such as downloads and citations (Mayernik, Hart et al., 2017).
然而, a number of recent studies have highlighted the limitations of studies that rely on cur-
rent data citation tracking infrastructures. Platforms such as DataCite have the potential to
enable large-scale studies of data production and its scholarly impact (例如, citations); 然而,
a lack of consensus on the definition of “data” and alignment of metadata across providers limits
DataCite’s analytical potential (Robinson-Garcia, Mongeon et al., 2017). An analysis of
publication-data set networks constructed from GenBank and Figshare found that authors tend
to cite publications over data sets, suggesting that historically, data sets have not been regarded
as first-class research objects and that data use inferred from citation networks may undercount
data use (曾, Wu et al., 2020). We avoid concerns about data and metadata quality by focus-
ing exclusively on a curated bibliography linking social science studies to publications held by a
single data archive. A recent study of ICPSR’s metadata records described the thematic and
temporal dimensions of social science data sets and their citing literature separately (李 & Jeng,
2019). We jointly analyze data and publications by constructing an interdisciplinary cocitation
网络. To tap the potential of shared data sets, we examine the role that data citations play in
the production and dissemination of knowledge in the social sciences.

2.3. Exclusive and Inclusive Communities in Knowledge Organizations

The analysis of citation networks can reveal hidden organizational structures. Cocitation anal-
ysis studies the structure of science and the emergence of specialties in bibliometric networks
by examining how frequently pairs of documents are invoked (小的, 1973). Author cocitation
analysis reveals individual contributions to specialty areas and paradigm shifts in the research
景观 (白色的 & McCain, 1998). Citation analysis can be used to identify exclusionary
community structures, such as “invisible colleges” (Price & Beaver, 1966)—in-groups that con-
trol scientific discourse, which are defined by strong ties and informal communication (Crane,
1977). Similar analyses can also detect “citation cartels” of authors who cite each other exclu-
sively, and effectively shut out other authors who work on the same subject (弗兰克, 1999). 在
addition to exclusionary practices, citation analysis can also identify convergence in research
社区. Studies of cross-field citation networks have found that fields of science tend to
become more integrated, rather than exclusive, over long periods of time (Varga, 2019), 甚至
incrementally across neighboring disciplines (Porter & Rafols, 2009).

Quantitative Science Studies

697

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
q
s
s
/
A
r
t
我
C
e
–
p
d

我

F
/

3
3
6
9
4
2
0
5
7
7
8
5
q
s
s
_
A
_
0
0
2
0
9
p
d

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

Subdivisions and crossroads

While the notion of “community” is central to these analytical methods, it is a difficult con-
cept to operationalize (Orthia, McKinnon et al., 2021); communities may take many forms,
and may play many roles. Identifying communities via data citation is further complicated
by the interdisciplinary nature of data analysis and citation (Heidorn, 2008). 然而, we take
inspiration from prior work showing that data reuse can be viewed as an indirect form of coop-
eration and collaboration between researchers—and groups that commonly reuse the same
data might be considered communities-at-a-distance (Sands et al., 2012; Thomer et al.,
2018; Zimmerman, 2008). Research data is a primary input for scientific knowledge produc-
的, making data archives important sites for identifying nascent research communities. 我们
use community detection to reveal patterns of data reuse and examine the structure of research
communities that use data as shared scientific inputs.

3. DATA AND METHODS

We analyzed the ICPSR Bibliography, an authoritative source of high-quality, manually
curated links between 8,071 social science studies and 101,674 publications that have cited
他们. An additional 2,420 学习 (23%) do not have any data-related publications and so are
not represented in the Bibliography. At ICPSR, each study consists of one or more data files and
metadata. 桌子 1 provides an example of available metadata for a highly cited ICPSR study.

Curation of the ICPSR Bibliography is labor-intensive, so the current coverage of the ICPSR
Bibliography is uneven6. Bibliography staff search broadly for academic literature that refer-
ences ICPSR studies and add literature to the Bibliography only if it analyzes ICPSR data or
includes an extensive discussion of data-related methodology. Publications in the Bibliography
are a mixture of materials published by the original data creator and publications that analyze
existing data. The majority of materials are journal articles, 报告, conference proceedings,
论文, 图书, and book chapters. We restricted our analysis to materials published since the
inception of ICPSR as an archive in 1962.

We analyzed citations for all of ICPSR’s currently available studies. Many ICPSR studies
have institutional principal investigators (PIs) including U.S. government agencies (例如, 我们.
Census Bureau, 司法部, 教育部, Department of Health and
Human Services), news outlets (CBS News, the New York Times), and university research cen-
特尔斯 (例如, University of Michigan’s Survey Research Center). Teams of individual researchers
also deposit data with ICPSR. Studies in our analysis included both restricted and public data
files. The terms of use for restricted data prohibit linking it to other data, so studies that include
restricted data may be undercounted in terms of their potential use.

The majority of ICPSR’s studies (62%) are also part of a series, meaning that they are part of
a recurring collection with new data archived over time (例如, repeated cross-sectional studies
or longitudinal studies). ICPSR provides access to 278 系列. We used a natural breaks clas-
sification (Jenks, 1963) to find highly cited series, which are reported in Table 2.

3.1. Network Definitions

We constructed citation networks from the ICPSR Bibliography, which are summarized in
桌子 3. Given that studies from the same series have been created intentionally to be

6 The process of retrieving citations for all studies is ongoing. Because staff are actively searching for publi-
cations that reference ICPSR data sets, these measures are minimum counts, which likely underestimate the
number of papers and their relationships.

Quantitative Science Studies

698

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
q
s
s
/
A
r
t
我
C
e
–
p
d

我

F
/

3
3
6
9
4
2
0
5
7
7
8
5
q
s
s
_
A
_
0
0
2
0
9
p
d

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

Subdivisions and crossroads

桌子 1.

Example of available metadata for an ICPSR study

Study name
Monitoring the Future:
A Continuing Study
of American Youth
(12th-Grade Survey), 1996

Series title
Monitoring the Future (MTF)
Public-Use Cross-Sectional
数据集

Release
1998-10-05

Citations
251

Subject terms

attitudes, demographic characteristics,
用药, family life, high school
学生, life plans, 生活方式, 社会的
行为, social change, 价值观, youths

analyzed together (例如, across years), we grouped studies by their series and referred to the
resulting unit as a “data set”—either one series with multiple studies or one study that is not
part of a series. Grouping studies into ICPSR-defined series allowed us to distinguish data that
were designed to be used together (例如, by their project sponsor, funder, archive) from data
that have been discovered to be useful together (例如, by researchers who cocite them in
文学).

Because publications and data sets are two different classes of objects in the ICPSR Bibli-
ography, we modeled the connections between them in a bipartite network (乙), consisting of
publication nodes, data set nodes, and edges linking publications to the data sets that they cite.
Citations are based on the total number of publications that use data from a study or series.
From network B, we projected data set nodes to create a weighted data set cocitation network
(S ). Edge weight in S indicates the total number of times that a pair of data sets have been used
together in publications. We removed low-frequency data cocitations from our analysis to
focus on data sets that were used together across multiple publications; we removed edges
from S with a weight less than 2, meaning that those data sets were only used together once.
This reduced edges by 87% (从 24,942 到 3,208) and nodes by 70% (从 3,363 到 998).

桌子 2.

Features of highly cited ICPSR series data

Series title
American National Election Study (ANES)

Lead investigators
Warren E. 米勒等人. and the National Election

Series

学习

学习
in series
92

组合
citations
16,771

Uniform Crime Reporting Program Data

Federal Bureau of Investigation

Series

Monitoring the Future (MTF) Public-Use

Lloyd D. Johnston et al.

Cross-Sectional Datasets

Current Population Survey Series

US Bureau of the Census

National Health and Nutrition Examination
民意调查 (NHANES) and Followup Series

Kathleen Mullan Harris et al.

National Survey on Drug Use and Health

United States Department of Health and Human

(NSDUH) Series

National Electronic Injury Surveillance

系统 (NEISS) Series

Services; 美国国立卫生研究院;
National Institute on Drug Abuse

United States Department of Health and Human
Services; Centers for Disease Control and
Prevention; National Center for Injury
Prevention and Control

National Crime Victimization Survey

Bureau of Justice Statistics

(NCVS) Series

263

296

Quantitative Science Studies

13,041

11,808

11,012

6,951

5,893

5,255

4,472

699

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
q
s
s
/
A
r
t
我
C
e
–
p
d

我

F
/

3
3
6
9
4
2
0
5
7
7
8
5
q
s
s
_
A
_
0
0
2
0
9
p
d

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

Subdivisions and crossroads

网络

Nodes

Edges

氮 (节点)

氮 (边缘)

Node size

Edge weight

成分

Density

Transitivity

Degree assortativity

桌子 3.

Summary of network definitions and metrics

乙

Publications, data sets

Data sets

Publication cited
ICPSR data set

ICPSR data sets cited in
the same publication

Fields of research

Publication tagged with
both fields cited one
ICPSR study

90,922 出版物;
3,363 data sets

998 data sets

129 research fields

102,580

持续的

n/a

1,687

2.3e−5

n/a

3,208

持续的

1 for each publication in

which the pair of ICPSR
data sets is cited

4,238

日志(Npapers)

1 for each ICPSR study
a publication cites

6.4e−3

0.28

−0.02

0.51

0.74

−0.30

下一个, we used a similar process to define a field of research network (F ) (坎宁安,
Smyth, & Greene, 2022). We gathered supplementary publication metadata for a subset of
44,639 publications in the ICPSR Bibliography (45% of the total) that were available in the
Dimensions database (Hook et al., 2018). We retrieved field of research (FoR) codes for each
出版物. FoR codes consist of 22 high-level divisions and their subgroups (例如, Curriculum
and Pedagogy is a subgroup of Education). We linked FoR codes to ICPSR data sets through
their corresponding publications in an unweighted bipartite network (B0). We then projected
the FoR nodes to create a weighted cocitation network (F ). 在F中, edges are data sets that are
cocited between fields of research. Because each study could be cited by many different com-
binations of fields of research, we did not group studies by their series, allowing for the obser-
vation of different cocitation patterns in the same series of studies. Edge weight indicates the
total number of times a pair of data sets have been used together in publications. We simplified
F by removing low-frequency FoR cocitations, which correspond to edges with a weight less
than five.

3.2. Community Detection

We applied community detection algorithms to each network as summarized in Table 4. 康姆-
munity detection identifies nodes that have a high probability of interacting based on the net-
work structure (Fortunato & Hric, 2016). We selected detection approaches based on the
desired representation of communities in each type of network (Lancichinetti & Fortunato,
2009; 哪个, Algesheimer, & Tessone, 2017). We allowed communities to overlap in the data
set cocitation network because we wanted to identify data sets with multiple roles. 然而,
we did not allow overlap in the field of research network because we wanted to find commu-
nities defined by members with the strongest ties.

Quantitative Science Studies

700

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
q
s
s
/
A
r
t
我
C
e
–
p
d

我

F
/

3
3
6
9
4
2
0
5
7
7
8
5
q
s
s
_
A
_
0
0
2
0
9
p
d

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

Subdivisions and crossroads

桌子 4.

Summary of community detection approaches

网络
S

Definition
Data sets (studies or series)

Community detection
方法
k-clique (k = 3)

Fields of research (FoR)

Louvain

in papers

Community definition
Data sets used in the same paper

Fields of research that use the

same study-level data

Communities
detected
41

We applied a k-clique percolation method to the data set cocitation network (S ) using the
corresponding implementation from the NetworkX Python library (Hagberg, Swart, & S Chult,
2008). A clique is a complete subgraph of a defined size (k) that can be reached from the
cliques of the same community through a series of adjacent cliques, meaning that the cliques
share k − 1 节点 (Palla, Derényi et al., 2005). Each node may belong to more than one clique,
resulting in overlapping communities. We selected a minimum clique size of three and labeled
each community with the three most common ICPSR subject terms for all studies in each
clique. Subject terms uniformly describe topics covered by the data and are defined by a con-
trolled vocabulary of social science concepts in the ICPSR Subject Thesaurus, 哪个是
assigned during data curation.

We then selected an aggregation-based method to represent communities in our field of
research network. We applied the Louvain algorithm to the FoR network (F ) using the corre-
sponding implementation from the Louvain Python library (Hagberg et al., 2008). 算法-
rithm uses modularity to discover communities in large networks by moving nodes locally to
create a network aggregation; communities are merged until the resulting modularity of the
overall partition can no longer increase (Blondel, Guillaume et al., 2008). This method results
in nonoverlapping communities that show the most densely connected fields of research that
cocite ICPSR data sets. The networks (S, F ) were then arranged with a spring layout, 哪个
places nodes with high degrees at the center of the graph.

4. 结果
We used two network measures—centrality and betweenness—to interpret the importance of
data sets and fields of research in their respective cocitation networks (纽曼, 2003). 第一的,
we calculated each node’s degree as the number of connections it shares with all other nodes
in the network. High-degree nodes are prominent in the network because they are highly con-
nected. We also calculated each node’s betweenness centrality by measuring all shortest paths
passing through a given node. Nodes with high betweenness function as hubs and connect
disparate parts of the network.

We also assessed structural features of the network—number of components, assortativity,
density, and transitivity—to compare the data set and field of research cocitation networks
(桌子 3). The FoR network is connected, meaning that all of its nodes are in the same com-
波南特, while the other two networks have multiple components or disconnected subgraphs.
This suggests that the FoR network is less complex than the data set cocitation network. Both S
and F exhibit negative degree assortativity, meaning that their nodes are less likely to be con-
nected to nodes in the network with a similar degree value. This pattern is stronger in F (−0.30)
than in S (−0.02). 最后, networks B and S have low density (2.3e−5 and 6.4e−3, 分别),
while network F is far denser (0.51), indicating that B and S have comparatively fewer edges
linking nodes and are not as easily traversed as F.

Quantitative Science Studies

701

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
q
s
s
/
A
r
t
我
C
e
–
p
d

我

F
/

3
3
6
9
4
2
0
5
7
7
8
5
q
s
s
_
A
_
0
0
2
0
9
p
d

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

Subdivisions and crossroads

4.1. Data Set Cocitations

The data set cocitation network (S ) has a periphery of data sets that have been used together
only a few times and a denser core of highly connected data sets, which are often used
一起. 数字 1 highlights important, central data sets, which are all found in the largest
subgraph at the core of the network. We used natural breaks to determine six data sets with
high betweenness and degree centrality, which play important roles in the network (桌子 5).

The important data sets we identified are long-running series made up of multiple studies.
Of these, the Uniform Crime Reporting Program Data Series has the highest degree and
betweenness. It has been used with 115 other data sets from studies or series across the citation
网络. The other data sets have strong ties to many other data sets and connect components
of the network. Half of these data sets are highly cited, with more than 10,000 citations each;
the others are less cited, yet play an important role in connecting the network. 最后, the lead
investigators for these important data sets include institutional PIs, meaning that one of the
study’s principal investigators or depositors is an institution (例如, the US Bureau of the Census),
and noninstitutional PIs.

To find collections of data sets that are often used together in publications, we performed
community detection on the data set cocitation network (数字 2). Not all studies belong to a
cocitation community. Only a fraction of data sets in the analysis (N= 632; 63%) belong to
cliques of size three or larger; these data sets are often analyzed with at least two additional
ICPSR data sets. The data sets that fell out of our analysis were used independently and were
not combined with other data sets. We labeled each community with the three most common
ICPSR subject terms for all data sets within it. The largest clique has 461 data set members and
is topically broad (例如, “demographic characteristics, 就业, income”) while smaller
cliques tend to have narrower focuses (例如, “terrorism, terrorists, radicalism”).

We also identified 20 data sets (3% of all nodes in the network) that belong to more than
one community, which may facilitate analyses across topics. Of these, we summarized data

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
q
s
s
/
A
r
t
我
C
e
–
p
d

我

F
/

3
3
6
9
4
2
0
5
7
7
8
5
q
s
s
_
A
_
0
0
2
0
9
p
d

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

数字 1. Overview of data set cocitation network featuring data sets functioning as hubs. Inset: High degree (红色的), high betweenness (蓝色的),
and high degree and betweenness (purple) 节点.

Quantitative Science Studies

702

Subdivisions and crossroads

桌子 5. Data sets with high betweenness and degree centrality in cocitation network

Data set name
Uniform Crime Reporting Program

Investigators
Federal Bureau of Investigation

Betweenness
0.17

Degree
115

Data Series

Studies in
系列
263

组合
citations
13,041

General Social Survey Series

National Opinion Research

中心; Davis et al.

American National Election

Study (ANES) Series

Miller et al.; National Election

学习

Current Population Survey Series

US Bureau of the Census

Census of Population and Housing,
1790–1950 [美国] Series

Haines et al.; US Bureau

of the Census

National Health Interview Survey

National Center for Health

Series

统计数据

0.12

0.11

0.10

0.05

113

109

117

296

155

1,551

16,771

11,012

818

4,448

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
q
s
s
/
A
r
t
我
C
e
–
p
d

我

F
/

3
3
6
9
4
2
0
5
7
7
8
5
q
s
s
_
A
_
0
0
2
0
9
p
d

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

数字 2. Result of community detection (41 communities detected at k = 3) with labels generated from the three most frequent subject terms
for the data sets in each community. An interactive graph with detailed node information is available in Tableau7.

sets that belong to more than two communities, along with examples of other data sets that
they have been cocited with, and a representative publication that has cited the same data
桌子 6. 例如, the Census of Population and Housing, 1790–1950 [美国]
Series appears in three different data set communities. It has been used with other ICPSR data
sets to study topics such as industrial development and urbanization in the United States; 骗局-
flict and international trade; and social movements and elections.

7 https://public.tableau.com/app/profile/ lizhou/viz/Study_communities_v2/Study_Communities_2

_Dashboard.

Quantitative Science Studies

703

问
你
A
n

t
我
t

桌子 6. Data sets in more than two communities, their cocited data sets, and publications

Data set
American National
Election Study
(ANES) Series

Community label terms
demographic characteristics,

就业, 收入

Example of cocited data sets

Example of citing publication

National Black Politics Study,

Wiegand, A. 瓦. (1999). Differences in public

[美国], 1993

public opinion, 政治的

attitudes, political behavior

Swedish Election Test-Data Series:
Swedish Election Study, 1979

opinion between blacks and whites: A social
psychological perspective. 大学
加利福尼亚州, Santa Cruz.

Granberg, D ., & Holemberg, S. (1991). 选举
campaign volatility in Sweden and the United
状态. Electoral Studies, 10(3), 208–230.

我

t
我
v
e
S
C
e
n
C
e
S
你
d
e
s

我

candidates, foreign policy,

American Representation Study,

爬坡道, K. Q., & Hurley, 磷. A. (1979). 大量的

national elections

1958: Candidate and
Constituent, Incumbency

Census of Population

demographic characteristics,

and Housing,
1790–1950 [团结的
状态] Series

就业, 收入

United States Agriculture Data,

1840–2012

international conflict, 战争,

Direction of Trade

nations

参与, Electoral competitiveness,
and issue-attitude agreement between
congressmen and their constituents. 英国人
Journal of Political Science, 9(4), 507–511.

Kitchens, C. T。, & Rodgers, L. 磷. (2020). 这
impact of the WWI agricultural boom and
bust on female opportunity cost and fertility
(不. w27530). National Bureau of Economic
研究.

McKeown, 时间. J. (1991). A liberal trade order? 这
long-run pattern of imports to the advanced
capitalist states. International Studies Quarterly,
35(2), 151–172.

census data, historical data,

19th century

National Samples from the Census
of Manufacturing: 1850, 1860,
和 1870

Dobis, 乙. A. (2016). The evolution of the American
urban system: 历史, 等级制度, and contagion.
Doctoral dissertation, Purdue University.

demographic characteristics,

Federal Justice Statistics Program

Bureau of Justice Statistics. (2021). Tribal crime

就业, 收入

Data Series

federal courts, sentencing,

defendants

Court Workforce Racial Diversity
and Racial Justice in Criminal
Case Outcomes in the United
状态, 2000–2005年

data collection activities. 技术报告. NCJ
301061, 华盛顿, 直流: Bureau of Justice
统计数据.

Ward, G。, Farrell, A。, & Rousseau, D. (2009). Does
racial balance in workforce representation yield
equal justice? Race relations of sentencing in
federal court organizations. Law & 社会评论,
43(4), 757–806.

sentencing, federal courts,

罪行

Impact of Sentencing Guidelines
on the Use of Incarceration in
Federal Criminal Courts in the
美国, 1984–1990

Tonry, 中号. (1991). Mandatory minimum penalties

and the US Sentencing Commission’s mandatory
guidelines. Federal Sentencing Reporter, 4(3),
129–133.

Monitoring of Federal
Criminal Sentences
Series

7
0
4

S
你
乙
d
我
v
我
s
我
哦
n
s

A
n
d

C
r
哦
s
s
r
哦
A
d
s

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
q
s
s
/
A
r
t
我
C
e
–
p
d

我

F
/

3
3
6
9
4
2
0
5
7
7
8
5
q
s
s
_
A
_
0
0
2
0
9
p
d

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

Subdivisions and crossroads

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
q
s
s
/
A
r
t
我
C
e
–
p
d

我

F
/

3
3
6
9
4
2
0
5
7
7
8
5
q
s
s
_
A
_
0
0
2
0
9
p
d

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

数字 3. Results of community detection in the field of research network (F, with nodes connected by edges of size ≥ 5). The interactive
graph with detailed node and edge information in size and study numbers is available in Tableau8.

8 h t t p s : / / p u b l i c . t a b l e a u . c o m / a p p / p r o f i l e / l i z h o u / v i z / C o m m u n i t i e s i n F i e l d s o f R e s e a r c h F o R

/CommunitiesinFieldsofResearch.

Quantitative Science Studies

705

Subdivisions and crossroads

4.2. Fields of Research

To find fields of research (FoR) that often use the same data sets, we performed community
detection on the FoR cocitation network (F ). Nodes in F are color-coded by their parent-level
divisions and labeled by their child-level code. We detected four large communities, 哪个
are summarized in Figure 3(A). The primary fields of research in each community are Human
社会, Philosophy and Education (社区 0); 经济学, Commerce and Management
(社区 1); Engineering, Earth and Environment, Information and Computer Science
(社区 2); and Medical and Health, 生物学 (社区 3).

Fields in the center of F have more cocitations, meaning that they are highly connected to
other fields. The central red frame in Figure 3(A) shows the major domains of research that cite
ICPSR data sets: Human Society, Philosophy and Education (社区 0). These central
domains are consistent with the idea that most items in the ICPSR Bibliography are social sci-
ence publications. 的确, social science (例如, Study of Human Society) and methodological
research fields (例如, 统计数据) are found in the core of the network while humanities and other
fields (例如, Creative Writing, Performing Arts) exist mostly on the periphery.

人物 3(乙)–(e) show the composition of each of the four communities in greater detail. 我们
found that the communities tend to divide along disciplinary lines. 例如, members
within each community are similar, in that they tend to share the same parent-level field of
研究. 例如, “Human Geography” and “Sociology” share the same parent-level
field of Human Society and are grouped into the same community (社区 0).

To examine the extent to which similar fields of research use the same data sets, we calcu-
lated citation statistics based on network F. We consider fields “similar” if they belong to the
same parent-level field (例如, “Civil Engineering” and “Environmental Engineering” are both
classified under Engineering) or the same community. We found that similar fields of research
cocite a limited range of data sets. The distribution of the aggregated numbers of data sets for
cocitation frequency by parent-level fields of research roughly follows a Poisson distribution
with λ = 1, indicating that as the number of parent-level fields citing the data set increases, 这
number of cocitations decreases (数字 4(A)). More than half (2,943 的 5,712) of the data sets
in F are cocited by only one community, further suggesting that data set use tends not to cross
community boundaries (数字 4(乙)).

We also observed core and periphery structures in the FoR network shown in Figure 3(A).
桌子 7(A) shows examples of fields of research located at the core of each community

数字 4. Data sets cited by parent-level fields of research. The y-axis indicates how many data sets were cited by the number of parent-level
fields on the x-axis. Most data sets are cited by a single parent-level field of research.

Quantitative Science Studies

706

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
q
s
s
/
A
r
t
我
C
e
–
p
d

我

F
/

3
3
6
9
4
2
0
5
7
7
8
5
q
s
s
_
A
_
0
0
2
0
9
p
d

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

Subdivisions and crossroads

桌子 7.

Examples of nonsocial science fields of research with core and periphery structures

(A) Fields in the core of each community subgraph

社区
membership
0

Field of research

心理学

认知科学

Law

应用数学

Library and Information Studies

Information Systems

统计数据

Artificial Intelligence and Image

加工

(乙) Fields in the periphery of each community subgraph

社区
membership

Field of research

Performing Arts and Creative

Writing

Curriculum and Pedagogy

Transportation and Freight

Services

Archaeology

Number of connected fields of
research—degree centrality of nodes
1,833

543

1,308

138

363

110

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
q
s
s
/
A
r
t
我
C
e
–
p
d

我

F
/

3
3
6
9
4
2
0
5
7
7
8
5
q
s
s
_
A
_
0
0
2
0
9
p
d

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

Example of frequently cocited data set and
corresponding fields of research

“National Crime Victimization Survey: School Crime Supplement,
2011”, cocited by fields including Policy and Administration,
Criminology, Sociology, Specialist Studies in Education,
心理学, Public Health and Health Services, 认知的
科学

“Midlife in the United States (MIDUS 2), 2004–2006”, cocited
by fields including Applied Mathematics, 银行业, Finance
and Investment, Economic Theory, Communication and
Media Studies, Business and Management, 政治学,
Econometrics, 应用经济学, Commercial Services

“American Time Use Survey (ATUS): Arts Activities, [美国],
2003–2018”, cocited by fields including Environmental Science
and Management, Other Economics, Religion and Religious
学习, Urban and Regional Planning, Information Systems,
Library and Information Studies

“National Health and Nutrition Examination Survey III, 1988–1994”,
cocited by fields including Anthropology, Demography, Clinical
科学, 统计数据, Artificial Intelligence and Image Processing,
Human Movements and Sports Science, 神经科学, Nutrition
and Dietetics, Other Medical and Health Sciences, Biochemistry
and Cell Biology

subgraph. They include a wide range of subfields such as Psychology, 统计数据, and Library
and Information Studies, which often advance methodological practices and make data-
related contributions. These nodes are highly connected to other fields of research and have
a much higher degree centrality compared to the average degree of nodes in F, 这是 9.

Quantitative Science Studies

707

Subdivisions and crossroads

Fields of research in the periphery of each community subgraph (桌子 7(乙)) reveal hidden
connections among disciplines through the data sets that they cocite. 例如, Archaeology
was cocited by 10 fields—while some of the cocitations are from social science disciplines such
as Anthropology and Demography, many others are related to biological and physical sciences,
including Clinical Sciences, 神经科学, and Nutrition and Dietetics, which are found in
社区 3.

5. 讨论

In this article we have applied metaphors from the built environment to interpret the hidden
research communities that we detected, and labeled the structures subdivisions and cross-
roads. These metaphors remind us that these communities of data use have emerged through
patterns of interaction in the research landscape and can be reshaped through intentional
设计. We refer to data sets in research as subdivisions if they are inward-facing, exclusive,
and not well connected to other data sets or fields. 反过来, we refer to data sets that are
often traversed by communities and fields as crossroads. We find 632 research data sets in
subdivisions that function as disciplinary resources; 20 research data sets at crossroads in
the network that function as boundary objects by facilitating interdisciplinary research; 和
nonsocial science fields that engage with social science data.

5.1. Subdivisions: Disciplinary Research Community Resources

We refer to data sets that serve a single disciplinary community as subdivisions because they
are inward-facing, exclusive, and not well connected. The largest data communities we
detected focus on international conflict, substance abuse, victimization, and public opinion
polls. Despite the topical breadth of the data set network (S ), it partitioned into coherent cli-
ques with a structure better described as a patchwork of subdivisions than a melting pot. 经过
比较, the FoR network (F ) had high density and high transitivity, suggesting that its
nodes tended to be clustered together. Given its cohesive structure, we partitioned F into a
small number of meaningful communities.

To understand the communities that function as subdivisions, we drew from a combination
of metrics computed for each network, which are summarized in Table 3. 全面的, the data set
cocitation network (S ) isn’t well connected. It has low density and low transitivity, is nonas-
sortative based on degree, and contains many components. 通过对比, the field of
research network (F ) has a negative degree assortativity, meaning that high-degree fields of
research nodes tend to attach to low-degree nodes. The network is not fractured compared
with the data set cocitation network (S ) and has only one component.

In the data set network (S ) shown in Figure 2, we found instances of isolated cliques with
data sets that were exclusively used together. 例如, we detected a clique of three data
sets described by the terms “Antebellum South (美国), slavery, slave labor.” These data sets
(“Southern Farms Study, 1860”; “Mortality in the South, 1850”; and “New Orleans Slave Sale
Sample, 1804–1862”) have different investigators and were produced for different purposes,
yet have been used together numerous times in academic articles. These three studies function
like a collection even though ICPSR did not designate them as one (IE。, by naming them a
系列). 一般来说, the analytic utility of data sets in subdivisions is limited to specific areas
of research. The notion of “thematic research collection”—a set of materials on a related
theme (Fenlon, 2017; 帕尔默, 2004)—may be useful for data archives to adopt; finding groups
of data used together is one way to identify candidate collections.

Quantitative Science Studies

708

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
q
s
s
/
A
r
t
我
C
e
–
p
d

我

F
/

3
3
6
9
4
2
0
5
7
7
8
5
q
s
s
_
A
_
0
0
2
0
9
p
d

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

Subdivisions and crossroads

We also found examples of cliques that shared topics, yet were disconnected from each other
(例如, “domestic violence, 罪犯, recidivism” and “domestic violence, 罪犯,
victimization”). While these data sets may be topically similar, researchers have not yet used
these data together. Cliques may be exclusive or disjointed for discovery reasons (IE。,
researchers outside of the user group are not aware of these data) or their data may be discov-
erable but unsuitable (例如, due to variables, 地理, or other properties). 例如, 一
community with data about “drug treatment” is composed of studies funded by the U.S. Depart-
ment of Health and Human Services, while a separate community of “substance abuse” data sets
is funded by the U.S. 司法部. These distinct communities may have stances
toward a research topic that are not interoperable and may even conflict.

In the field of research network (F ) 图中 3, we observed a subdividing tendency and an
in-group cocitation pattern for similar fields of research. These patterns of connection suggest that
each field of research cites a limited range of ICPSR data sets and supports the idea that ICPSR data
use divides along disciplinary boundaries (例如, social science disciplines such as economics and
education tend to cite the same data sets, but this is less common across nonsocial science fields,
such as engineering or nursing). Data sets in subdivisions have high analytic potential for narrow
communities of research; surfacing them and increasing their visibility may also help unlock
hidden potential for new uses beyond those narrow communities.

5.2. Crossroads: Engagement Across Research Communities

Data sets that facilitate interdisciplinary research are crossroads because they are often tra-
versed in connecting communities; in comparison to subdivision data sets, they are rare.
例如, ICPSR is well known for large series data sets (例如, American National Election
Study [ANES]), which attract data users to the archive. We found several of these series in the
largest clique (见表 5), which overlaps with the largest subgraph of the network. 这些
series are well known and have high engagement across multiple research communities. 在
特别的, the ANES Series and the Uniform Crime Reporting Program Data Series are institu-
tionally funded, highly cited, and connect a network of researchers who use them.

Prior work found a correlation between data sets with at least one institutional PI and higher
data reuse (Hemphill, Pienta et al., 2022). When we examine data reuse based on citations
rather than downloads, 然而, the relationship between data sets with institutional PIs and
reuse is less clear. Some institutional data sets already link multiple data sources into a single
data set and are useful on their own; they may not need to be combined with other data sets to
be analytically powerful. Among the crossroads data sets we found, the Census of Population
and Housing data set is unique because the individual investigator who constructed the data
set combined multiple years and data sources into a single data set, which has been broadly
useful across many applications.

In addition to the three connective data sets described in Table 6, we found 17 additional
data sets that function as crossroads between research communities. Many of these data sets
were often used with less cited data sets, explaining the negative associativity observed in net-
work F. 例如, the less cited “Vietnam Longitudinal Survey, 1995–1998,” is used with
the highly cited “India Human Development Survey (IHDS) Series” and “Chitwan Valley
[Nepal] Family Study Series” to study education, 家庭, and family planning. Researchers
who seek data from a well-known study may traverse the citation network to find complemen-
tary data sets from lesser-known studies. While a single data series such as the IHDS might
meet only some users’ needs, given its limited geographic coverage, the data sets linked
through its connections offer opportunities for comparative analysis.

Quantitative Science Studies

709

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
q
s
s
/
A
r
t
我
C
e
–
p
d

我

F
/

3
3
6
9
4
2
0
5
7
7
8
5
q
s
s
_
A
_
0
0
2
0
9
p
d

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

Subdivisions and crossroads

In the field of research network (F ), we found two dominant patterns of cocitation, 和-
marized in Table 7. Fields in the core of the network are highly connected and operate at an
interdisciplinary crossroads; they tend to use more data sets in common with other fields.
These fields, such as Statistics and Applied Mathematics, are not in the social sciences. 相当,
the data sets that they use function as crossroads, activating sites for research convergence. 在
社区 3 (数字 3(e)) 例如, Statistics cocites many of the same data sets as
生物学, 神经科学, and Medical Sciences. Statistical methods are often applied in data
analysis and can advance the development of methodologies in these areas. Fields on the
periphery of the network also seem to indicate new forms of engagement with social science
数据. 例如, the field of Transportation and Freight Services uses data from “American
Time Use Survey (ATUS): Arts Activities, [美国], 2003–2018,” along with Environ-
mental Science and Management, 经济学, Religion and Religious Studies, Urban and
Regional Planning, Information Systems, Library and Information Studies. 连接
between fields on the periphery of each community subgraph appear to maintain weak ties
among fields of research (Granovetter, 1973).

5.3. The Role of Research Data in Scientific Communities

These two structures suggest unique roles for data in scientific communities. Data sets in sub-
divisions and crossroads are two types of essential resources supporting social science
研究; subdivisions may have high disciplinary impact for the specific research domains
that use them, while data sets at a crossroads may provide connectivity across domains. 为了
实例, data at crossroads enable a kind of “arm’s length” cooperative work where the work
is loosely coupled, but depends on a “shared information space” that includes common data;
much like community efforts to maintain taxonomies across time and space, the shared
analysis of data sets contributes to cooperative “conceptual infrastructures” of scientific knowl-
边缘 (Bannon & 施密特, 1989, p. 361; Thomer et al., 2018).

While most ICPSR data is used by many disciplines within the bounds of social science,
data reuse outside of the social sciences tends to engage with data in two main ways. 第一的,
fields such as statistics and artificial intelligence are central in the field of research cocitation
网络; these fields may reuse social science data to develop new research and analytic
方法. 第二, fields such as performing arts and creative writing are peripheral in the
网络; while they tend to reuse ICPSR’s data less overall, they may provide novel inroads
for “awakening” cross-disciplinary data reuse in new application areas (胡 & Rousseau,
2019).

Identifying hidden communities and their structures within the data citation graph helps us
understand how data promotes knowledge production (Buneman et al., 2021; Lowenberg
等人。, 2019). It is likely that data sets occupying these different structures offer different types
of “analytical potential.” Palmer et al. (2011) describe “analytic potential” as “possible analytic
contributions for the range of possible user communities” (p. 4), and our method exposes those
possible communities and their structure. Research communities are beginning to recognize
the importance of contributing to data resources, and the citation graph enables us to assign
credit for different kinds of contributions (Alter & 冈萨雷斯, 2018; Cousijn et al., 2019).
Naming these different structures provides an accessible, extensible language for discussing
the functions of data and assigning credit for their creation. Creating and sharing data that
are used widely within one’s discipline ought to afford researchers credit among their peers,
sometimes for facilitating disciplinary depth—subdivisions—and at other times for creating
multidisciplinary resources—crossroads.

Quantitative Science Studies

710

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
q
s
s
/
A
r
t
我
C
e
–
p
d

我

F
/

3
3
6
9
4
2
0
5
7
7
8
5
q
s
s
_
A
_
0
0
2
0
9
p
d

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

Subdivisions and crossroads

5.4. Limitations and Outlook

Our analysis relied on a hand-curated resource, the ICPSR Bibliography of Data-Related
Literature, which limits the generalizability of our findings. While other data archives also
maintain bibliographies, few are as comprehensive and cross-cutting as ICPSR’s, 这是
maintained by dedicated staff who capture instances of data reuse across a wide variety of
media types and scientific disciplines. 然而, our network analysis method is generally
applicable to study data reuse and highlights incentives for data archives to maintain compre-
hensive bibliographies, which support the long-term study of data impact.

We also used the Dimensions database’s existing classification scheme for fields of
研究. This was a pragmatic choice given that codes were assigned at the level of publica-
tions rather than journals. 然而, the granularity of fields of research may be too coarse for
interpreting finer disciplinary patterns of data use within domain archives. Adopting other
domain analysis approaches could enhance our understanding of scientific knowledge pro-
归纳法 (Hjørland & Albrechtsen, 1995). 此外, we could compare the reuse of curated
social science data at ICPSR to self-archived data (例如, from the Dutch National Centre of
Expertise and Repository for Research Data: DANS).

We were able to identify data sets that served different purposes within scientific commu-
实体, but our data do not allow us to comment on how credit for creating different types of
data resources ought to be awarded to data creators and providers. Future research should
examine the relationship between data creation, reputation, and careers to understand how
to recognize data creators’ contributions. Because of the different roles they play in connecting
and supporting scholarly communities, data creators who produce subdivisions or crossroads
likely deserve different types of credit for their contributions. 例如, creating a data set
that operates as a subdivision should afford data creators substantial credit within their
discipline, while creating crossroads may award creators a broader reputation that is less
well-recognized within a single discipline. Data creators’ academic careers depend on how
they receive credit for their work and could impact the types of data resources they create
and share.

Our data is essentially a snapshot in time, and they do not enable us to investigate the
processes of community formation. The ICPSR Bibliography is a dynamic database; new cita-
tions are added continuously as they are discovered. The fact that a study does not have any
citations, or has very few, does not mean that its data have never been used; 相当, it may
mean that any existing records of its use have not yet been discovered. More exhaustive
searches for references to ICPSR data are under way. It is unclear whether data sets men-
tioned in literature (例如, “Data from the American National Election Survey (ANES) 是
restricted in its geographic coverage but contains valuable direct questions on the subjective
evaluation of racial groups …”) imply that the authors have analyzed the data or are men-
tioning the data for other purposes. Finer distinctions between types of data set references
will enable future studies of factors that contribute to the analytical potential and end-users’
decision to use data.

Given that data citation is a dynamic process, we are also interested in studying community
formation to better understand how social ties, data curation, or other factors shape data
citation networks. 例如, temporal citation dynamics provide rich insights into the
formation of research communities (Chubin, 1976). Extending the idea of “hibernation” to
research data sets that have not yet been “awakened” through reuse (胡 & Rousseau,
2019) and detecting bursts of citations following long periods of dormancy would allow us
to detect discovery events in the network. Understanding factors associated with novel data

Quantitative Science Studies

711

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
q
s
s
/
A
r
t
我
C
e
–
p
d

我

F
/

3
3
6
9
4
2
0
5
7
7
8
5
q
s
s
_
A
_
0
0
2
0
9
p
d

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

Subdivisions and crossroads

reuse would provide evidence to recommend underutilized research data and prioritize fund-
ing and credit for specific data curation activities.

6. 结论

Data citation networks contain hidden information about communities of data users and the
roles data play as primary inputs for scientific knowledge production. Through network anal-
分析, we revealed these communities and identified 41 communities of social science data sets,
along with four interdisciplinary research communities that use these data. Six important data
series connect the cocitation network. Data sets that are used together exclusively form
research subdivisions, which are valuable data collections for particular disciplines. 其他
data sets or fields that connect research communities are crossroads and have high topical
or analytical versatility. Research data sets that are produced for different purposes, 例如
long-running series data and single-purpose study data, are often used together. Similar fields
of research also tend to use the same combinations of data. 综上所述, these findings
contribute new ways of seeing scientific communities and make the impacts of research data
reuse visible.

致谢

Many thanks to Elizabeth Moss and the ICPSR Bibliography staff (Homeyra Banaeefar, Sarah
Burchart, and Eszter Palvolgyi-Polyak), David Bleckley, Elizabeth Yakel, Amy Pienta, 和
Dharma Akmon of the MICA team at the Inter-university Consortium for Political and Social
研究 (ICPSR) for their support of this research. We are also grateful to Sagar Kumar and
Andrew Schrock for providing feedback on our earlier drafts.

作者贡献

Sara Lafia: Analysis and interpretation of data, 概念化, 方法, 可视化,
Writing—original draft, Writing—review & 编辑. Lizhou Fan: Analysis and interpretation of
数据, 方法, 可视化, Writing—original draft. Andrea Thomer: 概念化,
资金获取, 监督, Writing—original draft. Libby Hemphill: 概念化,
资金获取, 方法, 监督, Writing—original draft.

COMPETING INTERESTS

The authors have no competing interests.

资金信息

This material is based upon work supported by the National Science Foundation under grant
1930645.

DATA AVAILABILITY

Citation data were derived from the ICPSR Bibliography in February 2022. Code for this
article’s analysis is available in a GitHub repository (Lafia, 2022乙) and data is available on
openICPSR (Lafia, 2022A). Access to licensed metadata from Dimensions was granted to
subscription-only data sources under a license agreement with Digital Science through the
密歇根大学.

Quantitative Science Studies

712

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
q
s
s
/
A
r
t
我
C
e
–
p
d

我

F
/

3
3
6
9
4
2
0
5
7
7
8
5
q
s
s
_
A
_
0
0
2
0
9
p
d

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

Subdivisions and crossroads

参考

Alter, G。, & 冈萨雷斯, 右. (2018). Responsible practices for data
sharing. The American Psychologist, 73(2), 146–156. https://土井
.org/10.1037/amp0000258, 考研: 29481108

Bannon, L. J。, & 施密特, K. (1989). CSCW: Four characters in
search of a context. ECSCW 1989. In Proceedings of the First
European Conference on Computer Supported Cooperative
工作. https://www-ihm.lri.fr/~mbl/ ENS/CSCW/2012/papers
/Bannon-ECSCW-89.pdf

Blondel, V. D ., Guillaume, J.-L., 兰比奥特, R。, & Lefebvre, 乙.
(2008). Fast unfolding of communities in large networks.
Journal of Statistical Mechanics: Theory and Experiment,
2008(10), P10008. https://doi.org/10.1088/1742-5468/2008/10
/P10008

Borgman, C. L。, Scharnhorst, A。, & Golshan, 中号. S. (2018). Digital
data archives as knowledge infrastructures: Mediating data shar-
ing and reuse. Journal of the Association for Information Science
and Technology, 70(8), 888–904. https://doi.org/10.1002/asi
.24172

Bose, R。, & Frew, J. (2005). Lineage retrieval for scientific data pro-
cessing: 一项调查. ACM Computing Surveys, 37(1), 1–28. https://
doi.org/10.1145/1057977.1057978

棕色的, C. (2003). The changing face of scientific discourse: Anal-
ysis of genomic and proteomic database usage and acceptance.
Journal of the American Society for Information Science and
技术, 54(10), 926–938. https://doi.org/10.1002/asi.10289
Buneman, P。, Christie, G。, 戴维斯, J. A。, Dimitrellou, R。, Harding,
S. D ., … Wu, 是. (2020). Why data citation isn’t working, 以及什么
to do about it. Database: The Journal of Biological Databases and
Curation, 2020, baaa022. https://doi.org/10.1093/databa
/baaa022, 考研: 32367113

Buneman, P。, Dosso, D ., Lissandrini, M。, & Silvello, G. (2021).
Data citation and the citation graph. Quantitative Science Studies,
2(4), 1399–1422. https://doi.org/10.1162/qss_a_00166

陈, C. (2017). Science mapping: A systematic review of the
文学. Journal of Data and Information Science, 2(2), 1–40.
https://doi.org/10.1515/jdis-2017-0006

Chubin, D. 乙. (1976). State of the field the conceptualization of sci-
entific specialties. The Sociological Quarterly, 17(4), 448–476.
https://doi.org/10.1111/j.1533-8525.1976.tb01715.x

Cousijn, H。, Feeney, P。, Lowenberg, D ., Presani, E., & Simons, 氮.
(2019). Bringing citations and usage metrics together to make
data count. Data Science Journal, 18(1), 9. https://doi.org/10
.5334/dsj-2019-009

Cousijn, H。, Kenall, A。, Ganley, E., Harrison, M。, Kernohan, D ., ……
克拉克, 时间. (2018). A data citation roadmap for scientific publishers.
Scientific Data, 5, 180259. https://doi.org/10.1038/sdata.2018
.259, 考研: 30457573

Cragin, 中号. H。, 帕尔默, C. L。, 卡尔森, J. R。, & Witt, 中号. (2010). 数据
sharing, small science and institutional repositories. Philosophi-
cal Transactions of the Royal Society A: Mathematical, Physical
and Engineering Sciences, 368(1926), 4023–4038. https://土井
.org/10.1098/rsta.2010.0165, 考研: 20679120

Crane, D. (1977). Social structure in a group of scientists: A test of the
“invisible college” hypothesis. In Social Networks (PP. 161–178).
爱思唯尔. https://doi.org/10.1016/B978-0-12-442450-0.50017-1
坎宁安, E., Smyth, B., & Greene, D. (2022). Navigating mul-
tidisciplinary research using field of study networks. In Complex
网络 & Their Applications X (PP. 104–115). https://doi.org
/10.1007/978-3-030-93409-5_10

Darden, L。, & Maull, 氮. (1977). Interfield theories. Philosophy of

科学, 44(1), 43–64. https://doi.org/10.1086/288723

Data Citation Synthesis Group. (2014). Joint declaration of data
citation principles. Force11. https://doi.org/10.25490/a97f-egyk
Fenlon, K. (2017). Thematic research collections: Libraries and the
evolution of alternative digital publishing in the humanities.
Library Trends, 65(4), 523–539. https://doi.org/10.1353/lib.2017
.0016

Fenner, M。, Crosas, M。, Grethe, J。, 肯尼迪, D ., Hermjakob, H。, ……
克拉克, 时间. (2016). A data citation roadmap for scholarly data
repositories. Scientific Data, 6, 28. https://doi.org/10.1038
/s41597-019-0031-8, 考研: 30971690

Fortunato, S。, & Hric, D. (2016). Community detection in networks:
A user guide. Physics Reports, 659, 1–44. https://doi.org/10.1016
/j.physrep.2016.09.002

弗兰克, G. (1999). Scientific communication—A vanity fair? 科学,
286(5437), 53–55. https://doi.org/10.1126/science.286.5437.53
Gökalp, 我. (1987). On the dynamics of controversies in a borderland
scientific domain: The case of turbulent combustion. Social
Sciences Information, 26(3), 551–576. https://doi.org/10.1177
/053901887026003005

Granovetter, 中号. S. (1973). The strength of weak ties. 美国人
Journal of Sociology, 78(6), 1360–1380. https://doi.org/10.1086
/225469

Gregory, K., Groth, P。, Scharnhorst, A。, & Wyatt, S. (2020). Lost or
成立? Discovering Data needed for research. Harvard Data
Science Review, 2(2). https://doi.org/10.1162/99608f92.e38165eb
Hagberg, A。, Swart, P。, & S Chult, D. (2008). Exploring network
结构, dynamics, and function using NetworkX. In Proceedings
of the 7th Python in Science Conference (SciPy) (PP. 11–15).
https://www.osti.gov/biblio/960616

Heidorn, 磷. 乙. (2008). Shedding light on the dark data in the long
tail of science. Library Trends, 57(2), 280–299. https://doi.org/10
.1353/库.0.0036

Hemphill, L。, Pienta, A。, Lafia, S。, Akmon, D ., & Bleckley, D.
(2022). How do properties of data, their curation, 和他们的
funding relate to reuse? Journal of the Association for Information
Science and Technology, 73(10), 1432–1444. https://doi.org/10
.1002/asi.24646

嘿, T。, Tansley, S。, & Tolle, K. (2009). The fourth paradigm: 数据-

intensive scientific discovery. Microsoft Research.

Hjørland, B., & Albrechtsen, H. (1995). Toward a new horizon in
information science: Domain-analysis. Journal of the American
Society for Information Science, 46(6), 400–425. https://doi.org
/10.1002/(SICI)1097-4571(199507)46:6<400::AID-ASI2>3.0
.一氧化碳;2-是

Hook, D. W., Porter, S. J。, & Herzog, C. (2018). Dimensions: 建筑
context for search and evaluation. Frontiers in Research Metrics
and Analytics, 3, 23. https://doi.org/10.3389/frma.2018.00023
胡, X。, & Rousseau, 右. (2019). Do citation chimeras exist? 案子
of under-cited influential articles suffering delayed recognition.
Journal of the Association for Information Science and Technol-
奥吉, 70(5), 499–508. https://doi.org/10.1002/asi.24115

Jenks, G. F. (1963). Generalization in statistical mapping. Annals of
the Association of American Geographers, 53(1), 15–26. https://
doi.org/10.1111/j.1467-8306.1963.tb00429.x

国王, G. (1995). Replication, 复制. 聚苯乙烯: 政治学 &

政治, 28(3), 444–452. https://doi.org/10.2307/420301

Lafia, S. (2022A). ICPSR Bibliography Citation Network (二月
2022) [Data set]. Inter-university Consortium for Political and
Social Research (ICPSR).

Lafia, S. (2022乙). ICPSR/data-communities (Version v1.0.0) [Computer

软件]. https://doi.org/10.5281/zenodo.6799127

Quantitative Science Studies

713

我

D
哦
w
n
哦
A
d
e
d

F
r
哦
米
H

t
t

:
/
/

d
我
r
e
C
t
.

米

我
t
.

e
d
你
q
s
s
/
A
r
t
我
C
e
–
p
d

我

F
/

3
3
6
9
4
2
0
5
7
7
8
5
q
s
s
_
A
_
0
0
2
0
9
p
d

乙
y
G
你
e
s
t

哦
n
0
7
S
e
p
e
米
乙
e
r
2
0
2
3

Subdivisions and crossroads

Lancichinetti, A。, & Fortunato, S. (2009). Community detection
算法: A comparative analysis. Physical Review E, 80(5),
056117. https://doi.org/10.1103/PhysRevE.80.056117, 考研:
20365053

李, J。, & Jeng, 瓦. (2019). The landscape of archived studies in a social
science data infrastructure: Investigating the ICPSR metadata
记录. Proceedings of the Association for Information Science
and Technology, 56(1), 147–156. https://doi.org/10.1002/pra2.62
光, 乙. A。, Clarkson, G。, Shedden, K., & 纽曼, 中号. 乙. J. (2007).
Large-scale structure of time evolving citation networks. Euro-
pean Physical Journal B, 59(1), 75–83. https://doi.org/10.1140
/epjb/e2007-00271-7

Lowenberg, D ., Chodacki, J。, Fenner, M。, 肯普, J。, & 琼斯, 中号. 乙.
(2019). Open data metrics: Lighting the fire. Zenodo. https://土井
.org/10.5281/zenodo.3525349

Mayernik, 中号. S。, 哈特, D. L。, Maull, K. E., & 韦伯, 氮. 中号. (2017).
Assessing and tracing the outcomes and impact of research infra-
结构. Journal of the Association for Information Science
and Technology, 68(6), 1341–1359. https://doi.org/10.1002/asi
.23721

Moss, E., & Lyle, J. (2018). Opaque data citation: Actual citation
practice and its implication for tracking data use. https://
deepblue.lib.umich.edu/handle/2027.42/142393

美国国家科学院. (2005). Facilitating interdisciplinary
研究. 美国国家科学院出版社. https://doi.org/10.17226
/11153

纽曼, 中号. 乙. J. (2003). Mixing patterns in networks. 物理评论
乙, Statistical, Nonlinear, and Soft Matter Physics, 67(2 Pt 2), 026126.
https://doi.org/10.1103/PhysRevE.67.026126, 考研: 12636767
纽曼, 中号. 乙. J. (2004). Who is the best connected scientist? A
study of scientific coauthorship networks. 在E中. Ben-Naim, H.
Frauenfelder, & Z. Toroczkai (编辑。), Complex networks (PP. 337–370).
施普林格. https://doi.org/10.1007/978-3-540-44485-5_16

Orthia, L. A。, McKinnon, M。, Viana, J. N。, & 沃克, G. (2021).
Reorienting science communication towards communities.
Journal of Science Communication, 20(03), A12. https://doi.org
/10.22323/2.20030212

Palla, G。, Derényi, 我。, Farkas, 我。, & 笑话, 时间. (2005). Uncovering the
overlapping community structure of complex networks in nature
and society. 自然, 435(7043), 814–818. https://doi.org/10
.1038/nature03607, 考研: 15944704

帕尔默, C. L. (2004). Thematic research collections. 在S. 苏珊, 右.
Siemens, & J. Unsworth (编辑。), A companion to digital humani-
领带. 布莱克威尔. https://www.digitalhumanities.org/companion
/看法?docId=blackwell/9781405103213/9781405103213.xml
&chunk.id=ss1-4-5&toc.depth=1&toc.id=ss1-4-5&brand
=default. https://doi.org/10.1002/9780470999875.ch24

帕尔默, C. L。, 韦伯, 氮. M。, & Cragin, 中号. H. (2011). The analytic
potential of scientific data: Understanding re-use value. 会议记录
of the American Society for Information Science and Technology,
48(1), 1–10. https://doi.org/10.1002/meet.2011.14504801174
Pasquetto, 我. 五、, Randles, 乙. M。, & Borgman, C. L. (2017). 上
reuse of scientific data. Data Science Journal, 16, 8. https://土井
.org/10.5334/dsj-2017-008

Porter, A. L。, & Rafols, 我. (2009). Is science becoming more interdisci-
plinary? Measuring and mapping six research fields over time. 科学-
tometrics, 81(3), 719. https://doi.org/10.1007/s11192-008-2197-2
Price, D. J. de Solla, & Beaver, D. (1966). Collaboration in an invis-
ible college. American Psychologist, 21(11), 1011–1018. https://
doi.org/10.1037/h0024051, 考研: 5921694

Robinson-Garcia, N。, Mongeon, P。, Jeng, W., & Costas, 右. (2017).
DataCite as a novel bibliometric source: Coverage, strengths and

局限性. Journal of Informetrics, 11(3), 841–854. https://土井
.org/10.1016/j.joi.2017.07.003

Sands, A。, Borgman, C. L。, Wynholds, L。, & Traweek, S. (2012). 福尔-
low the data: How astronomers use and reuse data. 会议记录
of the American Society for Information Science and Technology,
49(1), 1–3. https://doi.org/10.1002/meet.14504901341

小的, H. (1973). Co-citation in the scientific literature: A new mea-
sure of the relationship between two documents. Journal of the
American Society for Information Science, 24(4), 265–269.
https://doi.org/10.1002/asi.4630240406

Star, S. L。, & Griesemer, J. 右. (1989). Institutional ecology, “transla-
tions” and boundary objects: Amateurs and professionals in
Berkeley’s Museum of Vertebrate Zoology, 1907–39. Social
Studies of Science, 19(3), 387–420. https://doi.org/10.1177
/030631289019003001

Tenopir, C。, Allard, S。, Douglass, K., Aydinoglu, A. U。, 吴, L。, ……
Frame, 中号. (2011). Data sharing by scientists: Practices and
perceptions. PLOS ONE, 6(6), e21101. https://doi.org/10.1371
/journal.pone.0021101, 考研: 21738610

Thomer, A. K. (2022). Integrative data reuse at scientifically signif-
icant sites: Case studies at Yellowstone National Park and the La
Brea Tar Pits. Journal of the Association for Information Science
and Technology, 73(8), 1155–1170. https://doi.org/10.1002/asi
.24620

Thomer, A. K., Twidale, 中号. B., & Yoder, 中号. J. (2018). Transforming
taxonomic interfaces. Proceedings of the ACM on Human-
Computer Interaction, 2(CSCW), 1–23. https://doi.org/10.1145
/3274442

Tomasello, 中号. 五、, Vaccario, G。, & Schweitzer, F. (2017). 数据-
driven modeling of collaboration networks: A cross-domain
分析. EPJ Data Science, 6(1), 22. https://doi.org/10.1140
/epjds/s13688-017-0117-5

Varga, A. (2019). Shorter distances between papers over time are due
to more cross-field references and increased citation rate to
higher-impact papers. 美国国家科学院院刊
Sciences of the United States of America, 116(44), 22094–22099.
https://doi.org/10.1073/pnas.1905819116, 考研: 31611374
Vertesi, J。, & Dourish, 磷. (2011). The value of data: considering the
context of production in data economies. 在诉讼程序中
ACM 2011 Conference on Computer Supported Cooperative
工作 (PP. 533–542). https://doi.org/10.1145/1958824.1958906
白色的, H. D ., & McCain, K. 瓦. (1998). Visualizing a discipline: 一个
author co-citation analysis of information science, 1972–1995.
Journal of the Association for Information Science and Technol-
奥吉, 49(4), 327–355. https://doi.org/10.1002/(SICI)1097-4571
(19980401)49:4<327::AID-ASI4>3.0.一氧化碳;2-瓦

Wilkinson, 中号. D ., Dumontier, M。, Aalbersberg, 我. J。, Appleton, G。,
Axton, M。, … Mons, 乙. (2016). The FAIR Guiding Principles for
scientific data management and stewardship. Scientific Data, 3,
160018. https://doi.org/10.1038/sdata.2016.18, 考研:
26978244

哪个, Z。, Algesheimer, R。, & Tessone, C. J. (2017). A comparative
analysis of community detection algorithms on artificial net-
作品. Scientific Reports, 7, 46845. https://doi.org/10.1038
/srep46845, 考研: 28650447

曾, T。, 吴, L。, Bratt, S。, & Acuna, D. 乙. (2020). Assigning credit to
scientific datasets using article citation networks. Journal of Infor-
指标, 14(2), 101013. https://doi.org/10.1016/j.joi.2020.101013
Zimmerman, A. S. (2008). New knowledge from old data: The role
of standards in the sharing and reuse of ecological data. 科学,
技术, & Human Values, 33(5), 631–652. https://doi.org/10
.1177/0162243907306704