Talk:G-index

This article is rated Start-class on Wikipedia's content assessment scale. It is of interest to the following WikiProjects:

Academic Journals This article is within the scope of WikiProject Academic Journals, a collaborative effort to improve the coverage of Academic Journals on Wikipedia. If you would like to participate, please visit the project page, where you can join the discussion and see a list of open tasks.Academic JournalsWikipedia:WikiProject Academic JournalsTemplate:WikiProject Academic JournalsAcademic Journal articles See WikiProject Academic Journals' writing guide for tips on how to improve this article.

Perhaps I'm mistaken, but it looks to me that the alternative definition is just the h-index and should be removed here.... --Crusio (talk) 07:34, 12 September 2008 (UTC)[reply]

It would be great if someone could add a concrete, illustrative example with some simple numbers. This would make the article much more accessible. --Herr-Schlauschlau (talk) 19:56, 28 November 2011 (UTC)[reply]

...the bs-index? The so-called bullshit index?

It appears that the diagram for g-index is wrong. How can each column in the graph be an average? The g-index for the data shown in the diagram should be far higher than 11! The g-index is just the square root of the area under the citation curve (provided the author has that many publications). -- Dr. Kavi Mahesh (talk) 08:59, 23 July 2012 (UTC)[reply]

The X-es are the individual citation counts. The solid bars are the cumulative averages. It's not very clear, but the graph is correct. -- Luke Somers

The g-index is inappropriate for people who are minor participants in good papers early in their careers, in a way that the h-index is not. This happens from time to time, and it would be good for the index not to gag in such a case.

My citation counts are: 127, 40, 40, 14, 2, 2, 0. My h-index is 4, and my g-index is 7. As you might guess, I wasn't first or last author on that first paper. I did contribute to each of them meaningfully, and was a pivotal contributor to a majority of them (no, not the bottom 4). If I were to somehow publish an unlimited number of utter-dreck papers no one would ever cite, my g-index would rise to 15.

If one of my papers that's sitting around with 2 citations gets 3 more, then my h-index rise to 5, while my g-index remains unchanged (even with unlimited dreck available). It would take 27 more citations to raise my g; if these were distributed reasonably well my h-index would rise to 7.

Raising my h-index to 7 will realistically require writing 3 good papers, or perhaps writing 2 good papers and finding that my newer 2-cite paper hasn't peaked yet. Either route makes sense. On the other hand, raising my g-index presently requires writing a paper. Any paper. And this will continue to be the case for the next 8 papers even if no one ever cites me again.

This insensitivity to the down-list quality is a bug, not a feature. The g-index axiom (that a citation migrating from a weaker paper to a stronger one should not lower the index) is not a reasonable criterion for an index of academic performance on the low and mid ends. Perhaps it's valid for people who can be counted on to have been the driving force behind any smashing successes they were a part of.

-- Luke Somers — Preceding unsigned comment added by 24.218.57.142 (talk) 03:09, 15 May 2013 (UTC)[reply]

There is a big problem with saturation of the g-index. The usual definition of g-index implies that:

(1) The value of g may not exceed the total number of cited works.

However, this presents a problem since in this case g can saturates at the total number of works even as the number of citations increases to astronomical values. Saturation occurs whenever the average citation rate for all works exceeds the total number of works, which is all too common. It is a weakness because an exceptionally high citation rate is not fully counted, and the author is under-credited by the g-index measure. This is particularly a problem for the g-index, whose purpose is to avoid the kind of saturation effects one often finds in the h-index. As a way of over-coming this weakness, there is an alternate definition:

(2) If the average citation rate exceeds the number of works, then one continues the summation by padding the citation series with works having zero citations until the inequality criterion is satisfied.

For example, say a researcher has 5 works with the following citation counts: 40, 30, 20, 7, 3. The sum of the citations is 100. The average citation rate is 20. The average of the top-cited N-cited articles goes like (rounding to whole numbers):

N=1, Average=40

N=2, Average=35

N=3, Average=30

N=4, Average=24

N=5, Average=20

So g=5 would be the g-index according to case (1). But if we pad the series with zeroes like: 40, 30, 20, 7, 3, 0, 0, 0, 0, 0 ... then we would continue the running average of top-cited papers as follows:

N=6, Average=17

N=7, Average=14

N=8, Average=13

N=9, Average=11

N=10, Average=10

So according to case (2) the author has a g-index of 10. And if the author's works gained 10 more citations, she/he would go to g=11. And so on. Case (2) is a more natural way of defining g-index and taking into account high citation rates, and more closely follows the reasoning for having a g-index in the first place. — Preceding unsigned comment added by 131.112.112.6 (talk) 06:12, 21 December 2016 (UTC)[reply]