Page 131 - Academic and Research Book_WEB

Results can be skewed by self-citations: Although some self-citation is legitimate, authors can
profusely cite their own work to improve their h-index. This is particularly true in the case of large
collaborations involving hundreds of authors. Such collaborations churn out large number of
publications, which surely includes, in a natural manner, self-citations. It is natural that all authors
in these collaborations would have large h-index, which actually reflects the contribution of the
collaboration and not of the individual authors.

Results aren’t comparable across disciplines: Citations depend on a particular subject and the sub-
category or work area. The h-index varies widely by subject. A mediocre h-index in the life sciences
will still be higher than a very good h-index in the social sciences. Comparison across disciplines is
fraught with dangers, and at best be avoided. We can’t benchmark h-indices because they are rarely
calculated consistently for large populations of researchers using the same method.

Results can’t be compared between researchers: The h-index of a researcher with a long publication
history including review articles cannot be fairly compared with a post-doctoral researcher in the
same field, nor with a senior researcher from another field. Researchers who have published several
review articles will normally have much higher citation counts than other researchers.

Unable to capture an unusual original work: In the final analysis, the h-index would miss such a
work, which many times come to the mainstream research after a gap. Further, the index does not
distinguish between who is citing and where. For example, few citations by acknowledge leaders of
the field are more important than large numbers general citations. Similarly, few citations in quality
journals are far better than large number of citations by all and sundry working in the field.

There are several related metrics that are derived from h-index. For example, H-core relates to the
set of top cited h-articles from a journal, whereas h-median represents median of citation counts in a
journal’s h-core. One can restrict the citation window to the preceding five years to check the output of
an author in those five years. The corresponding H-index is referred to as h5, similarly we can define
h5-core and h5-median, corresponding to articles published in the five years.

Leo Egghe’s g-index

In 2006, Leo Egghe suggested another important author level index, the g-index (Egghe, 2006). The
index is calculated based on the distribution of citations received by a given author’s publications.
Suppose the research papers are arranged in decreasing order of the number of citations that these
have received, then the g-index is the unique largest number such that the top g-articles have received
together at least g citations. It can be equivalently defined as the largest number g of highly cited
2
articles for which the average number of citations is at least g. Effectively, the g-index by design jacks
up the low-cited papers by the highly-cited papers.

Comparing h-index and g-index

In the case of h-index, once a paper has sufficient number of citations to get included in the h-index,
additional citations to the same article are not that important, essentially implying the papers that
have much higher citations than the h-index do not matter for that particular value. In other words
h-index defines essentially a quality threshold of the publications. The g-index, in contrast, weighs
highly-cited papers more heavily. The g-index value will always be equal to or greater than the h-index
value.

120 | University Grants Commission