Klaas Wynne
17 April 2009
Hire at a normalised 1 only
Tags: citations, h-index, hiring, professors, publications, ResearcherId, web of sciencePosted in High-impact journals, Tips
The other day I met a professor from a pretty good department and he said something along the lines of “of course the h-index is not important but it is funny that all the hires we have made over the years had a normalised h-index of 1 or above and when they didn’t, questions were asked”. So the conclusion really is that to be hired you need a normalised h-index greater or equal than 1.
What’s this all about? First the h-index, in case you have been in the bush and haven’t heard about it yet. There’s good explanation on the wiki page on the h-index. Briefly, if you order your publications by the number of citations in decreasing order and index them 1,2,3, etc., when the index becomes equal to the number of citations, you have found your h-index. Thus, an h-index of 10 means that you have 10 papers with 10 citations or more. It filters for outliers such as one paper with a 1000 citations. The normalised h-index (hbar?) is the h-index divided by the number of years that have passed since your first publication. Supposedly (I can’t find where I read this now), an h-index of 20 when you have been in your career for 20 years qualifies you to become a full professor.
Anyway, the professor I mentioned above probably referred to hires at a junior level (assistant professor, lecturer, whatever it is called where you are). It might be reasonable to assume that junior hires have been in a postdoc for about 5 years, so his department would expect and h-index of 5 after 5 years. Does this sound reasonable to you? I personally think that having a look at somebody’s h-index is good as long as you dig a bit deeper to find out why it is low or high and look at other things too.
Speaking of h-indices. Web of Science now connects to something called ResearcherId where you can set up web pages for yourself displaying your h-index and graphs showing the number of citations per year. A colleague of mine (different from the one above) with a normalised h-index of about 1.7 has such a page and I thought you might be suitably bowled over by his citation graph. Suffice to say that my graph doesn’t look quite as impressive…
Follow
17 Apr 2009 13:33, Allard Mosk
Apparently, the h-index does quite a good job as a single-number indicator of performance. However, there is something about it that worries me a lot: It is used often by physicists, who essentially believe that a unit that is useful today (the Watt, for example) will still be useful tomorrow. However, the h-index really is a concept that belongs in political economics, and, as we all know since recently, in that science units which were useful yesterday may be worthless today.
As the h-index gains popularity, scientists will work to optimize their h-index, not their raw output (which is not a one-dimensional quantity), as they can gain jobs, money, influence by optimizing h-index. It is like a bonus in the banking world. So, when people start basing salaries or even hiring decisions on h-index, in a number of years (I think two or three rather than 10) the h-index will mostly be an indicator of how good you are at optimizing your own h-index.
The conclusion is, any single number that characterizes performance should only ever be used historically. I guess this is the research area of economics, or evolution biology. Are there any experts from those fields that can comment?
18 Apr 2009 17:04, Otto Muskens
The normalised h-index as you describe as starting from your first publication is ill-defined. Many people are co-author during their undergraduate thesis, which would place the starting point for counting unrealistically early in their career. Further on you mention the end of your PhD as a starting point, which is at least a more clearly defined but usually lags several years behind the first publication of PhD related work. So in order to use the normalized h-index, it should be well-defined and fair, i.e. not penalizing early output. Otherwise, the absolute h-index is a better measure than the normalized one.
19 Apr 2009 21:25, Bram van Ginneken
I agree with Allard’s conclusion, but I find the h-index a useful figure. Its main virtue is simplicity. And it’s actually not so one-dimensional: you should ask people to provide not just the number but the list of h papers (with number of citations per article) that make up their h-index. That’s typically a fairly short list. Glancing this list you quickly see if the papers are recent, if the author substantially contributed to the papers (or was always somewhere in the middle of the author list), if the top papers have many more citations than h, if the papers cover a range of topics, if the list does not contain too many review papers, etc.
Once your h-index gets larger, optimizing it becomes rather difficult, I think (but if somebody knows a really good trick, let me know 😉 )
21 Apr 2009 15:34, jacopo
Hey! My normalized h-index is bigger than one! Can I apply for a position at the “pretty good department”? 🙂
Ok, silly jokes apart, just like all scalar descriptor it is obvious that h-index alone cannot grasp the complexity of a subject like “scientific success”. Nevertheless it contains quite e big deal of information and, in conjunction with other data, can help to formulate an opinion on someone career.
In the original paper from Hirsch he proposed to use the normalized h-index as a way to describe the degree of scientific success of someone in his late career (in his calculation Anderson has a normalized h-index of 1.88 and Hawkings 1.41). Using this parameter to judge someone still at the beginning (like me) or even a “fresh” professor (or someone who is applying for a position) is prone to give paradoxical results.
Another (mayor) issue is that changing your field the average number of citation you get can vary drastically. As far as I know mathematicians are the most greedy and biologists the most generous when it comes to the number of references they place in one of their papers. Therefore a mathematician with a normalized h-index of 1 is (probably) far better than a biologist with the same score.
Incidentally this suggest a way to optimize the h-index. Just collaborate from time to time with someone in a field where the number of citations is much higher than in your field.
On this variability among different fields I’d like to point out this paper that describes a normalization procedure that might alleviate this problem.
8 Jun 2010 23:08, labuddy
I agree with what have been said above.
Should the normalization be done against the total number of publications he/she authored/co-authored on AND counted in calculating the h-index? In doing so, everyone will have a number between 0 and 1. E.g. one with an h-index of 70 who have indeed published 700 papers can be more meaningfully compared against one with an h-index of 30 who only published 200 papers.
That should also take away the “junior-authorship penalty”. Right? We can further take away the “age” consideration totally out of the picture.
Furthermore, this h-index should definitely be field-dependent. For example, some scientists by training are placed in engineering department to teach. They published in some more general area (basic chemistry for instance) than more focus journal (mechanical engineering for instance). These scientists’ hi-indexes could thus not be objectively compared with their engineering colleagues.
My 2 cents only!
24 May 2012 14:16, Why every scientist should make his Google Scholar profile public | CIENCIA
[…] profileYou want an example? Just check my Google Scholar Citations profileRelated posts:Hire at a normalised 1 onlyIncreasing your h-index by reclaiming misspelled citationsEvery scientist should have a Researcher […]