Page 106 - Journal of Library Science in China, Vol.45, 2019

CHEN Bikun, ZHOU Huixian, ZHONG Zhouyan & WANG Yuefen / Exploring the user platform preference and 105
user interest preference of Chinese scholarly articles: A comparison based on usage metrics

Different from discipline perspective, average downloads per article on journal official websites
is greater than CNKI in 1/3 CSSCI-indexed journals and 37/40 CSCD-indexed journals in the
perspective of journals, which indicates that users of social sciences generally prefer pay-for-
access information integration platform but users of natural sciences generally prefer open access
journal official websites to obtain required papers.
(2) Skewness analysis
From Table 2, it shows that average downloads per article in each discipline is higher than
its median, indicating that usage data are concentratedly distributed. In order to reveal the
concentration, skewness analysis is applied. In the first step, articles in each discipline are sorted
by descending number of usage counts. Next, functional relationship is calculated between
cumulative percentage of articles in each discipline and cumulative percentage of its usage counts.
The same steps are done for citations. The three functional relationships are plotted in Figure 3 and
4. They reveal that usage counts and citations are all skewedly distributed. Downloads of official
sites and CNKI citations are skewedly distributed in a similar way and they are more skewedly
distributed than CNKI downloads.
In the perspective of disciplines, top 25% papers in each discipline contribute more than 50%
downloads and about 60% citations. Specifically, in CNKI platform, downloads of “Library,
Information and Archival Science” and “Pedagogy” are mostly concentrated, then “Economics”,
“Computer Science”, “Biology”, “Management Science”, “Earth Science” and “Math”.
Citations of “Pedagogy” and “Math” are mostly concentrated, then “Computer Science” and
“Library, Information and Archival Science”, while “Economics”, “Biology”, “Earth Science”
and “Management Science” are the least. On journal official web sites, downloads of “Library,
Information and Archival Science” and “Earth Science” are mostly concentrated, then “Pedagogy”,
“Economics”, “Management Science” and “Biology”, and “Computer Science” and “Math”
are the least. Combined with descriptive statistics, it reveals that downloads of “Earth Science”
(professional discipline with the highest mean) on journal official sites are mostly concentrated,
then downloads of CNKI platform, and citations are the least. Downloads of “Math” (basic
discipline with the lowest mean) are least concentrated on journal official sites and CNKI platform.
But citations of “Math” are mostly concentrated. Downloads and citations of “Pedagogy” (applied
disciplines with middle mean) are all mostly concentrated.
In addition, tail of downloads (80%-100% in Figure 3 and 4) on journal official sites are lower
than CNKI platform in all disciplines, which means that tail of downloads was paid more attention
in CNKI platform than journal official websites. By analyzing distribution of head and tail of
downloads, it reveals that information integration platforms attract more users and offer each paper
more opportunity to be used and journal official websites show more obvious “Matthew Effect”
and more professional characteristics, confirming user preference differences between information
integration platforms and journal official sites.