Page 140 - JOURNAL OF LIBRARY SCIENCE IN CHINA 2018 Vol. 43

140 Journal of Library Science in China, Vol.9, 2017

3.2 Distribution features of the bibliometric model of usage

1) Absolute value distribution: approximate and positively skewed distribution at high-frequency
part
Similar to the far-from-rare phenomenon of zero-citation articles, the usage data also show that
each discipline has some zero-usage literature (see Table 2 and Figure 2). While zero-citation
literature can be explained with the aid of “sleeping beauty”, zero-usage literature seems to require
a new explanation. According to the relevant data, the proportions of zero-usage articles in both
economics, and Library and Information Science have slightly exceeded 1%, i.e., 1.05% and 1.13%,
respectively. The topic of zero-usage literature demands continuous attention from follow-up
literature. Meanwhile, the absolute usage count values of different disciplines have different high-
frequency zones. To be specific, the high-frequency usage count of physics is mainly concentrated
in the range of 1–12, in which 3 represents the peak frequency value. The usage count of computer
science is mainly concentrated in the range of 2–8 with a peak frequency value of 5. The peak
frequency value of the usage count of social sciences is higher than that of natural sciences. The
peak value of the usage count of economics is 9 whilst that of Library and Information Science is
12. According to the left part of Figure 2, the citation count falls between zero and peak value, and
the number of articles significantly decreases with the increase of usage count. According to the
right part of Figure 2, the absolute value distribution law of usage count is different from that of
citation count. As indicated by the results of fitting and measurement, in the high-frequency zone
of usage for the four disciplines (the usage count concentrated zone before and after citing the peak
value), the usage count and the number of articles present a normal distribution (the goodness of
the normal distribution fitting exceeds 0.9 in each case, but the charts actually reveal a positively
skewed distribution). Among the four disciplines, economics has a relatively standard distribution
pattern. When the K-S test is carried out on the absolute usage of all the literature for each of the
four sample disciplines, the results show that none of them conforms to normal distribution. The
reason for this can also be observed from the charts. The minimum value part of usage has a lower
number of pieces of literature, and the high-frequency part declines at an excessively fast pace,
making it impossible to show a normal distribution. Clearly different from the significant absolute
value power-law distribution of citation, the absolute value distribution of citation is neither a
highly standard power-law distribution nor a qualified normal distribution. Instead, the absolute
value distribution only shows local positively skewed distribution in the high-frequency value
phase. Given that power-law distribution is a common distribution pattern of bibliometrics (Egghe,
2005; Zhao & Ye, 2013), this absolute value distribution feature of usage once again suggests that
usage may be able to play a special role in the academic evaluation theory.
2) Cumulative value distribution: approximate power-law distribution in a cumulative integral.
According to the data of this paper, the usage data of the WoS platform show no unified law
in terms of absolute value distribution. However, the test shows that the cumulative value of the