Table 2.
Interdisciplinarity measures reviewed in this study
MeasureFormula
p_multi The percentage of multi-assigned journals
p_outside The percentage of journals that are classified in more than one research area
pro kicik/∑jcij
d_links The number of links between different SCs established by journals in a given category
Pratt index $2n+12−∑kgpikn−1$, where g is the index obtained by ranking pik in decreasing order
Spec k$cik2$/(∑kcik)2
Simpson index 1 − ∑k$pik2$
Shannon entropy −∑kpik lnpik
Brillouin index (log(∑kcik)! − ∑k log cik!)/∑kcik
Gini coefficient $∑k2h−n−1cikn∑kcik$, where h is the index attained by sorting SCs according to cik in increasing order
RS j.k(1 − sjk)pijpik
Hill-type measure 1/∑j.ksjkpijpik
Coherence j,k$cjki$(1 − sjk
BC1 j,k$rjikrjk$
CC jPj$cijaiaj$
AS iPi ($1N$jsij), where N is the number of all other SCs
MeasureFormula
p_multi The percentage of multi-assigned journals
p_outside The percentage of journals that are classified in more than one research area
pro kicik/∑jcij
d_links The number of links between different SCs established by journals in a given category
Pratt index $2n+12−∑kgpikn−1$, where g is the index obtained by ranking pik in decreasing order
Spec k$cik2$/(∑kcik)2
Simpson index 1 − ∑k$pik2$
Shannon entropy −∑kpik lnpik
Brillouin index (log(∑kcik)! − ∑k log cik!)/∑kcik
Gini coefficient $∑k2h−n−1cikn∑kcik$, where h is the index attained by sorting SCs according to cik in increasing order
RS j.k(1 − sjk)pijpik
Hill-type measure 1/∑j.ksjkpijpik
Coherence j,k$cjki$(1 − sjk
BC1 j,k$rjikrjk$
CC jPj$cijaiaj$
AS iPi ($1N$jsij), where N is the number of all other SCs

1Leydesdorff (2007) used a symmetrical cosine matrix instead of a citation matrix to measure the interdisciplinarity of journals using BC. He claimed that the size of journals was “controlled” in this way. However, some researchers still use a citation matrix to measure BC (for instance Silva et al., 2013). In our view, it is more intuitive to use a citation matrix as input for BC, as it better demonstrates the concept of intermediation. Hence, we use a citation matrix in the present study. The shortest path indicates the path with the lowest total edge weight. However, in our case, results obtained by BC can be scaled with size (i.e. if SC i has a large number of publications, it is likely to have a high degree of interdisciplinarity when BC is used). BC was calculated using the R package SNA: Tools for social network analysis. The input citation matrix was generated using SQL from the in-house WoS database at CWTS. Other measures were calculated using SQL or a combination of SQL and R.

Close Modal