Models for author per paper distribution
Steven. A. Morris
Steven.A.Morris at OKSTATE.EDU
Fri May 20 09:36:39 EDT 2005
I am looking for models for the author per paper distribution of
literatures. I would be especially interested to participate in a
discussion about this with anyone who has looked into the topic..
I have done quite a bit of searching in Scientometrics and related
journals, but the authors per paper distribution is usually discussed as
a side topic in papers about collaboration It is kind of hard to find
such asides, since the literature on collaboration is so huge.
My idea is that "little science" specialties, , i.e., specialties that
don't exhibit "hyperauthorship", exhibit an authors per paper
distribution that approximates a 1-shifted Poisson distribution. This
is the same as saying that the distribution of the number of secondary
authors on papers is Poisson distributed. I am interested in finding
evidence in the literature that either supports or refutes the
"1-shifted Poisson" claim. I am aware that there are some specialties,
like high energy physics, that have authors per paper distributions with
power-law tails. Those are outside of the scope of what I'm looking at.
I've found only two papers [1,2] that discuss the author per paper
distribution in detail:
Beaver asserts that the secondary authors per paper distribution is
Poisson, and while not presenting any data as evidence, states that the
Poisson fits especially well to specialties up until the beginning of
1900, but that as collaboration increased through the 20th century, the
tail of the distribution tended to deviate, while the body of the
distribution remained firmly Poisson. Beaver's justification of the
Poisson distribution is made from random networks theory- in a network
of papers connected by randomly occuring common author links the
distribution of degree of links would be a Poisson distribution. In
a later paper, (more like notes from a presentation), Beaver reassserts
that the distribution is Poisson and states that some fields are moving
toward a power law distribution for authors per paper.
Ajiferuke et al, in a very nice 1988 paper which was mentioned on this
list last week, presents a table of authors per paper from Library
Science Abstracts and discusses several possible distributions,
1-shifted Poisson, 1-shifted binomial, and more. Ajiferuke's data, when
plotted, approximates the 1-shifted Poisson distribution pretty well,
though this is not discussed in the paper. Ajiferuke's interpretation
of the mechanism generating the number of secondary authors is also
believable: the number of secondary authors is number of researchers
that the lead author consults with before the completion of paper. The
usual arguments justifying a Poisson queuing process can be believably
applied to such an interpretation.
Among other papers on collaboration, Seglen and Aksnes show a plot of
authors per paper in the field of microbiology and assert that the
distribution well matches a 1-shifted Poisson distribution. I digitized
the data from the figure in the paper and replotted it here
It seems to me that their data does indeed well-approximate a 1-shifted
In another paper, Glanzel presents data from three fields, mathematics,
chemistry, and biomedical research. Dr. Glanzel was kind enough to send
me the original data from this study which is plotted here
seems to me that this data also, shows that the author per paper
distribution well approximates a 1-shifted Poisson distribution, though
there appears to be a definite trend in biomedical sciences toward a
distribution with a fatter tail.
My own observations, based on data from 27 case studies covers a wide
variety of fields is shown here
<http://samorris.ceat.okstate.edu/web/author_dist/default.htm>. In many
cases the data fits the 1-shifted Poisson distrbution very well, in
other cases, particularly biomedical specialties, the fit is not that good.
If anyone out there has any comments or ideas about the topic, I'd be
very happy to hear from them.
Oklahoma State U.
 D. deB. Beaver, "Teamwork: A Step beyond Collaboration", George Sarton
Centennial, Communication and Cognition, Ghent, Belgium, (1984) 449-452
Ajiferuke, I. , Burrell, Q. & Tague, J. (1998). Collaborative
>coefficient : A single measure of the degree of collaboration in
>research. Scientometrics, 14(5-6), 421-433.
Steven A. Morris, Ph.D steven.a.morris at okstate.edu
Electrical and Computer Engineering office: 405-744-1662
202 Engineering So. mobile: 405-269-6576
Oklahoma State University
Stillwater, Oklahoma, 74078, U.S.A
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the SIGMETRICS