this is for holding javascript data
Alberto Pepe edited tori and riq.tex
about 11 years ago
Commit id: efa04ab7290bde92631c3e935cf509e371082504
deletions | additions
diff --git a/tori and riq.tex b/tori and riq.tex
index 2adda33..bbefbd0 100644
--- a/tori and riq.tex
+++ b/tori and riq.tex
...
the proposed measure employs two more bits of bibliographic information,
readily available by modern scholarly databases: the
number of authors and the number of references in a paper.
Both these measures have been used before, separately. Adjusting
citation counts for the number of authors seems obvious
system since 1996 \cite{kurtz05b}. Adjusting for the number of
references has become a standard technique in evaluating journals,
with Web of Science using Eigenfactor \cite{west} and SCOPUS using SNIP
indices have been proposed in the literature, as discussed above. For example, dividing
the \textit{h}-index by the number of authors in a paper
discipline \cite{radicchi} both yield promising results for
cross-disciplinary impact comparison.
Thus, we normalize every external (non-self) citation received by a scholar
in two ways: by the number of authors in the cited paper and by the
number of references in the citing article. We speculate that {\it a
simple double normalization, by number of authors and by number of
references in the citing article, has the effect of grounding
productivity index in the authorship and citation practices of a given
field at a given time.}
We define the Total Research Impact, {\it tori}, of a scholar as:
\begin{equation}
tori = \displaystyle\sum_{n} \frac{1}{a \cdot r}
\end{equation}
where $n$ is the collection of external (non-self) citations accrued
by the researcher, $a$ is the number of authors of the cited paper,
and $r$ is the number of bibliographic references of the citing
paper. One calculates the overall, cumulative output of a scholar by
summing the impact of every external citation accrued in his/her
career. As such, the total research impact of a scholar ({\it tori}) is
simply defined as \textit{the amount of work that others have devoted
to his/her research, measured in research papers}.
The definition of {\it tori} influences the self-citation correction. The
standard self-citation correction \cite{Wuchty} removes a citation
if any of the authors of the citing paper are the same as the authors
of the cited paper. With the computation of {\it tori}, we only remove a citation {\it if the
author being measured is an author of the citing paper.} \cite{glanzel}
We can also compute the research impact averaged over a scholar's
career, equivalent to the \textit{m}-quotient. For a scholar with a
career span of $y$ years, the Research Impact Quotient, \textit{riq},
is defined as:
\label{formula:ror2}
\begin{equation}
riq = \frac{\sqrt{tori}}{{y}}
\end{equation}
We test the performance of this measure on the same corpus discussed
above, finding that the research output quotient performs very well
both over time and across sub-disciplines of astronomy, as shown in
Figures 3 and 4.
Temporal debasement effects are greatly attenuated when computing the
perform, on average, similar to astronomers who started publishing 50
years later (global mean is $x = 0.098$). An exponential best-fit
regression line (shown as solid line, with a $0.95$ confidence band)
still shows a positive gradient ($b = 0.00659$), but considerably
smaller than that of {\it m}. (A similar analysis on a cohort of 544
astronomy Ph.D. confirmed this result, finding an exponential
regression line with slope $b = 0.00417$). The large attenuation of
temporal effects obtained with the computation of {\it riq} is not
predominately due to either of the two normalizations: they both
contribute roughly equally. The temporal slope after removing the
effects of multiple co-authors is $b = 0.015$ and the slope after the
normalization by number of references only is $b = 0.020$.
The disciplinary bias observed for the \textit{m}-quotient, previously
discussed and depicted in Figure 1, are greatly
improved when the \textit{riq} is computed, as shown in Figure
do perform below (i.e., ``gravitational waves'' and ``cosmic rays'')
or above average (i.e., ``stars parameters'', ``galaxies clusters'', and
``particles''), the lower and upper {\it riq} quartile band
measured for the majority of
fields analyzed (18 out of 23) tends to fall within the global mean