A Critique of the Proposed National Education Policy Reform
Groups of Highly Cited Publications: Stability in Content with Citation Window Length
1. Vienna, Austria, 15-19 July 2013, http://www.issi2013.org/
Introduction
Groups of highly cited publications may change in content of individual
publications with extending citation windows, while remaining stable in size.
Such content changes can be relevant in particular when highly cited publications
are used to assess the level of excellence for publication records of limited size,
e.g. of individual scientists.
This poster paper shows content changes between groups of highly cited articles
identified in different citation window lengths, in 2 domains with different
citation characteristics: a Physics sub-domain (faster) and Mathematics (slower).
Groups of Highly Cited Publications:
Stability in Content with Citation Window Length
Nadine Rons, Nadine.Rons@vub.ac.be, http://rd-ir.vub.ac.be/en_GB/people/show/id/554
Research Coordination Unit, Vrije Universiteit Brussel (VUB), Pleinlaan 2, B-1050 Brussels, Belgium
Specifications and notations
Data: Data sourced from Thomson Reuters Web of Knowledge (formerly referred to as ISI Web of Science
- WoS). WoS accessed online, 01-02.2013. Document type: ‘article’ only, Publication year: 2004. Citation
windows: up to 8 years, including publication year.
Two domains, each consisting of three partition cells containing all journals assigned to a same
combination of subject categories[a]:
Physics sub-domain: Cells consisting of the journals assigned to subject category ‘Astronomy Astrophysics’
only (A, 8047 articles), to ‘Physics Particles Fields’ only (P, 1977 articles), and to both (A&P, 2440 articles).
Mathematics: Cells consisting of the journals assigned to subject category ‘Mathematics’ only (M, 10022
articles), to ‘Mathematics Applied’ only (MA, 3938 articles), and to both (M&MA, 3286 articles).
Highly cited publications: The frame below shows details for ‘outstandingly’ cited articles identified in
the third iteration of the parameter free Characteristic Scores and Scales (CSS) technique[b]. The poster
paper in the proceedings also discusses, in less detail, similar results obtained for ‘remarkably cited’
publications, and with pre-set percentiles of 1% and 5%.
References [a] Rons, N. (2012). Partition-based field normalization: An approach to highly specialized publication records. Journal of Informetrics, 6(1), 1-10.
[b] Glänzel, W. & Schubert, A. (1988). Characteristic scores and scales in assessing citation impact. Journal of Information Science, 14(2), 123-127.
[c] Glänzel, W. (2007). Characteristic scores and scales: A bibliometric analysis of subject characteristics based on long-term citation observation. Journal of Informetrics, 1(1), 92-102.
Changing CONTENT of groups of outstandingly cited articles with citation window length
Summary and conclusions
A group of highly cited publications does not contain exactly the same individual publications calculated in
citation windows of different length. As citation window length is gradually incremented per year,
successive groups of highly cited publications more strongly resemble in content [1.1], evolving towards a
phase of higher ‘content stability’ that is still subject to changes from year to year [1.2].
Characteristics of publications entering and leaving the group of highly cited publications with increasing
citation window length vary with domain, regarding their relative amounts [2.1] from year to year [1.3], and
citation levels before entering [2.3] or after leaving [2.2] the group of highly cited articles.
The changing content of a group of highly cited publications with citation window length is an important
issue in particular in methodologies for the assessment of publication records of limited size, such as
those of individual scientists.
For assessments of small subjects, effects of choices regarding citation window length deserve particular
attention in domains with ‘slow’ citation characteristics, and in comparisons between domains with
different citation characteristics.
[2] Shifts of individual articles
Change over 5 years, between citation
windows of 3 and 8 years
Between a small 3-year citation window
and a large 8-year citation window, change
in content can be substantial
[Vertical axis: logarithmic scale, value 1 represents 1 or 0 accumulated citations]
DOMAIN DEPENDENCE:
With increasing citation window length,
articles partly shift to other citation
levels, more so in Mathematics than in
Physics, in relative amount and in range
of citation levels:
(illustrated left for 1 cell per domain; numbers
below for all 3 cells per domain)
[2.1] The relative amount of articles entering
the group of outstandingly cited articles later
than the first 3 years (black dashed lines
versus colored full lines) is lower in Physics
(0,3 to 0,4) than in Mathematics (0,7 to 0,8).
[2.2] Most articles leaving the group of
outstandingly cited articles between 3 and 8
years (grey dashed lines) are still remarkably
cited in the 8-year window, more often in
Physics (80-100%) than in Mathematics
(65-75%).
[2.3] Most articles entering the group of
outstandingly cited articles between 3 and 8
years (black dashed lines) already are
remarkably cited in the 3-year window, more
often in Physics (79-94%) than in
Mathematics (58-78%).
[2.4] The most cited articles in the 3-year
and 8-year citation windows are the same in
Physics, and are different in Mathematics.
Occurrence of articles changing
outstandingly cited status in publication
records of individual scientists:
(preliminary observations, not illustrated)
In the observed cells, a number of individual
scientists published more than one article in
2004 that gained outstandingly cited status
later than the first 3 years:
2 articles for 19 scientists in Physics,
representing 7% up to all of the scientist’s
articles from 2004 in this domain;
2 to 4 articles for 16 scientists in
Mathematics, representing 11% up to all of
the scientist’s articles from 2004 in this
domain.
[1] Overall content resemblance
Changes between consecutive citation
windows, incremented by one year
Between two citation windows, one year
different in length, changes in content
become smaller between larger windows
[1.1] Groups of outstandingly cited articles identified in
two consecutive citation windows have an increasing part
of content in common for increasing citation windows. A
similar convergence in content is observed in each cell of
the two domains. (illustration above: larger intersections
for pairs of consecutive citation windows of higher length)
[1.2] In sufficiently large citation windows, a later more
stable phase with smaller remaining changes in content
from year to year is reached.
[1.3] DOMAIN DEPENDENCE:
A same degree of ‘content stability’ between groups
of outstandingly cited articles identified in
consecutively incremented citation windows is
reached sooner in Physics than in Mathematics.
While their content changes, the groups of outstandingly
cited articles remain stable in size with extending citation
windows beyond an initial citation period of about 3
years[c] [frame below].
Stable SIZE of groups of highly cited articles
(illustrated below for Characteristic Scores and Scales (CSS): stability beyond an
initial citation period of about 3 years, see also [c]; obvious for pre-set percentiles)