Page 109 - Journal of Library Science in China, Vol.45, 2019

108 Journal of Library Science in China, Vol.11, 2019

From Figure 5, it depicts that the overall distribution of social science and natural science is
similar and Figure 5a and 5c reveals more obvious stratification in different disciplines than
Figure 5b and 5d. Figures 5a and 5c show negative correlation and Figures 5b and 5d show
positive correlation (when log transformation of CNKI downloads is not calculated, the linear
regression coefficients will be higher, but when log transformation of website downloads is not
calculated, the linear regression coefficients are still very low). Similarly, Zhao (2017) showed
relatively higher correlation coefficient between citations and downloads in “Physics” and “Library,
Information and Archival Science” and lower correlation coefficient in “Computer Science” and
“Economics” (significance test has passed). Specifically, linear regression coefficients of social
science and natural science are both low (even negative in “Math”) in journal official websites.
In CNKI platform, linear regression coefficients of social sciences (Figure 5b) are higher than
natural sciences (Figure 5d), and coefficients of each discipline in social sciences are similar, and
coefficients of “Earth Science”, “Computer Science” and “Biology” in natural sciences are similar,
but coefficient of Math is lower.
In a word, downloads and citations are not correlated on journal official websites but correlated
on CNKI platform. Combined with functions provided by CNKI platform, it is found that papers
with higher downloads or citations are listed on the top of default search results pages, which
attracts users to download and read “top papers”. But there are not similar functions on journal
official websites. Users download papers with their original needs and judgement. Therefore, the
above results reveal that user patterns of downloads and citations are affected by unique functions
of different platforms.
(4) Correlation analysis of website downloads and CNKI downloads
In order to further reveal relations of website downloads and CNKI downloads, Spearman
coefficient of each discipline and journal is calculated in Table 3 and 4. From Table 3, only
“Library, Information and Archival Science” is moderately correlated, and other disciplines are not
correlated or negatively correlated. From Table 4, only journal “Chinese Journal of Biologicals” is
positively correlated, and 14 journals (in bold font) are moderately correlated. Overall, correlation
between website downloads and CNKI downloads is low, which indicates that users’preferences
for various platforms are quite different.

Table 3. Spearman coefficient of CNKI downloads and website downloads in different disciplines
Disciplines Correlation Disciplines Correlation
Library, Information and Archival Science 0.406 *** Computer Science 0.087 ***

Management Science 0.324 *** Earth Science -0.013
Economics 0.035 Math -0.219 ***
Pedagogy 0.065 * Biology 0.316 ***
(*p<0.05; **p<0.01; ***p<0.001)