Page 191 - Becoming Metric Wise

182 Becoming Metric-Wise

difference in citations between year 6 and year 7 is 1251 1051 corre-
sponding to a difference in percentage points of 9.36. The difference
between 49.2 (actually 49.16) and 50.00 is 0.84. Hence we need another
0.84/9.36 of a year to reach 50%. This is 0.09 years. Hence the median
age of cited articles of this journal is 6.09 or rounded to one decimal 6.1
year. This is how cited and citing half-lives are calculated in the JCR.
Consider now the slightly changed Table 6.14.
Yet one could make another point. The journal represented by
Table 6.14 has a median cited age of exactly 6. What is the average age of
the 198 articles published in the year 2007 and cited in the year 2012?
One would expect them to be, on average, 5 years old. Indeed, if all cita-
tions were received on the last day of the year 2012 and if all cited articles
were published on the first day of the year 2007 than these articles would
be almost 6 years old. If however, all citations were received on the first
day of the year 2012 and all cited articles were published on the last day
of the year 2007 then these articles would be slightly more than 4 years
old. On average such citations refer to articles that are 5 years old
(Rousseau, 2006b). So, the median cited age (as calculated in the JCR)
refers to articles that are on average half a year less old than this median
cited age. We think that it would be more logical if the time line as pre-
sented in the JCR were shifted over one half year. The point we want to
make here is that an article cited in the year Y (here 2012) and published
in the year Y 2 5 (here 2007) is not, on average, 5 1/2 years old as sug-
gested by the second row of Table 6.10 (assuming a uniform distribution
of citations over the year), but only five (on average), as explained above.
Even then, there is still a problem for the first year, but for this technical
issue we refer the reader to (Rousseau, 2006b).
One of the points of discussion about the Garfield-Sher JIF is that the
numerator takes all citations into account (received within the citation win-
dow) while the denominator only counts “citable documents.” This leads to
two problems: the first is that the notion of a “citable document” is not pre-
cisely defined (although it always includes “normal articles” and reviews),
and the second one is that in some journals “noncitable documents” receive
quite a lot of citations (for instance some editorials or letters to the editor).
In this context such citations are sometimes referred to as “for-free-
citations.” We return to this point in the sections on impact factor manipu-
lation (see Section 6.15) and on CiteScore (Section 6.16).
The JCR provides the data used to calculate the indicators shown in its
database. This includes, for instance, the number of citable publications and