Page 148 - Becoming Metric Wise

139
Publication and Citation Analysis

5.10.3 Applications of Bibliographic Coupling and
Cocitation Analysis

Cocitation analysis and bibliographic coupling can be applied to many
types of actors: authors, journals, countries (via their authors’ affiliations),
and so on. The most-studied type of cocitation analysis is Author
Cocitation Analysis or ACA in short. ACA is most often used to analyze
the intellectual structure of a scientific field. It was introduced by White
and Griffith (1981). In 1990 Kate McCain published a technical overview
of ACA, which became a standard for this particular application
(McCain, 1990). She distinguishes four steps:
• In the first step one constructs the raw cocitation matrix (expressing
how often two authors are cocited during a particular citation win-
dow, in a given database).
• Next this matrix is transformed into a proximity, an association or a
similarity matrix.
• In the third step one applies multivariate statistical analysis. The tech-
nique applied in this step is often MDS (multidimensional scaling),
cluster analysis, factor analysis or correspondence analysis (statistical
techniques not studied in this book). This results in a two-
dimensional map on which authors that are often cocited are repre-
sented in each other neighborhood. If this technique is successful
authors with similar research interest form groups.
• In the last step one proposes an interpretation of the results.
In early applications of this four-step approach Pearson’s correlation
coefficient was used to obtain a similarity matrix. However, Ahlgren,
Jarneving, and Rousseau (2003) have shown that it is possible that the
Pearson correlation coefficient may lead to results leading to the opposite
of what is desired (or what is logical). This observation is nowadays gen-
erally accepted leading to the use of Salton’s cosine measure instead of the
Pearson correlation coefficient. Moreover, Leydesdorff and Vaughan
(2006) and later Zhou and Leydesdorff (2016) pointed out that one
should start from the citation matrix and directly obtain a normalized
cocitation matrix, where normalization can be performed by the cosine
similarity.
Bibliographic coupling is used in a similar way as cocitation analysis,
but is less popular. It seems that results are somewhat more difficult to
interpret (cf. step 4). Yet, the WoS applies bibliographic coupling in its
Related Records link (McVeigh, 2009).