Page 145 - JOURNAL OF LIBRARY SCIENCE IN CHINA 2018 Vol. 43

ZHAO Xing / Exploring the measurement features of usage data for academic literature 145

citation count and usage count in the sample data of this paper, the literature of the four disciplines
all show some statistically significant positive correlation with certain strengths (see Table 4),
supporting Proposition 2 empirically. In physics, as well as Library and Information Science,
the positive correlation between citation count and usage count is relatively strong. In computer
science and economics, the correlation is weak, but passes the significance test, suggesting the
presence of a statistical correlation. These results combine to suggest that, notwithstanding the
specific value or ranking differences, which are certainly significant, as regards the trends of value
and ranking, there is no dichotomy between usage and citation.

Table 4. Pearson and Spearman’s correlation analysis on the usage count and citation count of four disciplines
Discipline Pearson correlation coefficient Spearman correlation coefficient
Physics 0.635** 0.395**
Computer Science 0.304** 0.407**
Economics 0.345** 0.367**
Library and Information Science 0.415** 0.516**
**: p<0.01(two tail)

Table 4 also shows that the ranking correlation and value correlation of the two indexes are not
consistent, so analyzing the difference between the two from ranking also provides a beneficial
perspective. There are relatively significant differences between usage count and citation count in
the specific ranking. For instance, the mean ranking difference of physics based on the two indexes
reaches as high as 30,249, the ranking difference of single literature reaches as high as 129,208,
and only three out of 100,000 articles report a consistency between usage count-based ranking and
citation count-based ranking. In computer science, the mean ranking difference between usage
count and citation count is 9,990 with a maximum value of 41,816. There is no article showing
consistency between the two rankings. In many evaluation applications, the absolute value is not
very important and the key lies in the ranking in the discipline. The significant ranking differences
described above tell us that in actual applications, the results of usage have some independence
and can serve as an important supplement to citation.

3.4 Case analysis on the discipline of Library and Information Science

As indicated by the comparison between citation and usage as well as by their correlation analysis,
for a single academic article, the inconsistency between usage count and citation count is relatively
widespread. According to the analysis conducted by Wang, Fang, and Sun (2016) on five Library
and Information Science journals, in this field, the difference between citation data and usage data
is obvious. To discuss the specific embodiment of such a difference, this paper selects the literature
with quantitative features in usage and citation for analysis. Table 5 shows the literature ranked in