Johan Bollen, Marko A. Rodriguez, and Herbert Van de Sompel "Journal Status" arXiv:cs.GL/0601030 v1 9 Jan 2006

Stephen J Bensman notsjb at LSU.EDU
Mon Mar 6 12:41:27 EST 2006

I see that you have once again taken my name in vain and again given me the
opportunity to spout my ideas on SIGMETRICS.  I must admit that I have not
read the paper you discuss, because my doctor warned me against reading too
many such papers, since I am fairly close to OD-ing on them.  However, the
conclusions you mention do seem a little peculiar.

Due to detailed study of Gene Garfield's development and utilization of
impact factor, I am coming to change my mind on this measure somewhat.  It
is for rather complicated reasons, which I shall try to explain below.

In general I think that there is too much random error in citation data for
the utilization of such precise techniques as correlation--Pearson,
Spearman, whatever.  Much results from exogenous citations due to an
inability to define precise sets--a logical consequence of Bradford's Law
of Scattering and Garfield's Law of Concentration.  Impact factor suffers
from a further source of error due to an inability to classify precisely
sources into citable and non-citable--something which honest persons can
disagree on.  This inability severely affects the denominator of the impact
factor equation.  What is therefore needed is a technique that is crude and
robust against such error.  I have personally found it in the chi-square
test of independence, which allows the conversion of citation measures into
ordinal variables defined by broad categories.  It also allows one to
define the amount of error one is willing to accept, i.e., upper 10% vs.
upper 25%.

Use of this chi-square test may vindicate impact factor by demonstrating
that it has the same strong relationship to expert ratings as do total
citations.  As a matter of fact, it may be a superior measure in that it
will not only capture the importance of reseach journals but also of review
journals.  Close inspection of the top 10% of the journals recommended by
the LSU chemistry faculty reveals it to be a balanced mix of research
journals, review, journals, and the main teaching journal of chemistry.
In other words, most facets of journal importance are captured by this
measure, whereas total citations captures mainly research, and impact
factor captures chiefly the review journals.  However, broadening the
categories may cause impact factor to capture both research and review
though not the teaching facet.  In any case I am going to test this in the
revision of the JASIST paper I am now engaged in.

Impact factor has the ability to do this for the very reasons Seglen
denounces it.  His main case against is based on the reasoning of the law
of error and the role of the arithmetic mean in this law.  This requires
the normal distribution for the arithmetic mean to be an accurate estimate
of central tendency.  However, due to the highly skewed distributions with
which we deal, the arithmetic mean is always way above the other estimates
of central tendency such as the median or the geometric mean due to the
high degree of variance caused the dominant observations.  Seglen's
reasoning collapses once one realizes that a journal's or scientist's
importance is not measured by central tendency but by the variance caused
by the few important articles published by the journal or scientist.
Therefore, scientific importance is the result of variance and not central
tendency.  The arithmetic mean, which impact factor attempts to estimate,
better captures the variance.

To demonstrate, I have converted Garfield's constant for the year 1993 into
binomial p and the Poisson lambda  The way I did this is in the attached
Excel file.  You will see the binomial p is a lousy 0.0003, which converts
into a Poisson lambda or Garfield's constant of 2.15 for the year.  This is
the probability or the rate articles were cited in 1993 on the assumption
of probabilistic homogeneity.  However, since there is probabilistic
heterogeneity, most articles have to have a citation rate below Garfield's
constant.  True to form, of the 5000 journals covered that year, 4500
journals were below to Garfield's constant.  2.15 is an awful small range
to squeeze 4500 journals into and expect meaningful quantitative
distinctions.  Utilization of a central tendency measure puts one right
smack in the middle of that tight range.  Small as this may be, the
probabilities and lambda were actually much smaller, for Garfield's
constant is based on the set of articles actually cited that year, i.e., it
it truncated on the left and does not take into account the articles that
could have been cited but were not.  I do not have the technical or
intellectual ability to estimate this zero class.  I do know that Sir
Maurice Kendall backed off from the problem when he confronted it in
Bradford's Law, and who the hell am I compared to Maurice Kendall.  I wish
that somebody would write an article understandable to simpletons on how to
make such estimates.  From my perspective, this would be one of the most
important articles ever written.

Sorry for the tirade, but I thought I'd float a few trial balloons to be
shot down.


(See attached file: GarConst.xls)

Loet Leydesdorff <loet at LEYDESDORFF.NET>@LISTSERV.UTK.EDU> on 03/04/2006
07:14:57 AM

Please respond to ASIS&T Special Interest Group on Metrics

Sent by:    ASIS&T Special Interest Group on Metrics

cc:     (bcc: Stephen J Bensman/notsjb/LSU)

Subject:    Re: [SIGMETRICS] Johan Bollen, Marko A. Rodriguez, and Herbert
       Van de Sompel "Journal Status" arXiv:cs.GL/0601030 v1 9 Jan 2006

Dear colleagues,

The idea is interesting. However, there a few problems with this paper.
First, the authors should not have used Pearson correlation coefficients to
compare the rankings, but rank correlations (Spearman's rho or Kendall's
tau). Second, it would have been interesting to have a rank correlation
"total cites" given recent discussions (Bensman). Third, the delineation of
fields in terms of the ISI subject categories is very questionnable.

However, the authors are very clear about their results: "We identified ...
, but were unable to recognize a meaningful pattern in the results." (p.
I don't understand why one should then multiply the one measure with the
other. What does multiplication to the error?

Does one of you know a place where the ISI subject categories are
How are they produced? People seem to use them increasingly both in
evaluation and research practices, but I have never been able to reproduce
them using journal citation measures.

With best wishes,


Loet Leydesdorff
Amsterdam School of Communications Research (ASCoR),
Kloveniersburgwal 48, 1012 CX Amsterdam.
Tel.: +31-20- 525 6598; fax: +31-20- 525 3681;
loet at ;

> -----Original Message-----
> From: ASIS&T Special Interest Group on Metrics
> [mailto:SIGMETRICS at LISTSERV.UTK.EDU] On Behalf Of Eugene Garfield
> Sent: Friday, March 03, 2006 6:37 PM
> Subject: [SIGMETRICS] Johan Bollen, Marko A. Rodriguez, and
> Herbert Van de Sompel "Journal Status" arXiv:cs.GL/0601030 v1
> 9 Jan 2006
> Adminstrative info for SIGMETRICS (for example unsubscribe):
> Further to yesterday's posting, "Prestige is factored into
> journal ratings", here is another interesting and informative article
> email: {jbollen, marko, herbertv}
> TITLE   :  Journal Status
> AUTHORS :  Johan Bollen, Marko A. Rodriguez, and Herbert Van de Sompel
> SOURCE   : arXiv:cs.GL/0601030 v1 9 Jan 2006
> Abstract
> The status of an actor in a social context is commonly
> defined in terms of two factors: the total number of
> endorsements the actor receives from other actors and the
> prestige of the endorsing actors. These two factors indicate
> the distinction between popularity and expert appreciation of
> the actor, respectively. We refer to the former as popularity
> and to the latter as prestige. These notions of popularity
> and prestige also apply to the domain of scholarly
> assessment. The ISI Impact Factor (ISI IF) is defined as the
> mean number of citations a journal receives over a 2 year
> period. By merely counting the amount of citations and
> disregarding the prestige of the citing journals, the ISI IF
> is a metric of popularity, not of prestige. We demonstrate
> how a weighted version of the popular PageRank algorithm can
> be used to obtain a metric that reflects prestige. We
> contrast the rankings of journals according to their ISI IF
> and their weighted PageRank, and we provide an analysis that
> reveals both significant overlaps and differences.
> Furthermore, we introduce the Y-factor which is a simple
> combination of both the ISI IF and the weighted PageRank, and
> find that the resulting journal rankings correspond well to a
> general understanding of journal status.
> ______________________________________________
> Adminstrative info for SIGMETRICS (for example unsubscribe):
>    OR
> Philip Ball : p.ball at
> Title: Prestige is factored into journal ratings
> Author(s): Ball P
> Source: NATURE 439 (7078): 770-771 FEB 16 2006
> Document Type: News Item Language: English
> Cited References: 0      Times Cited: 0
> ISSN: 0028-0836

-------------- next part --------------
A non-text attachment was scrubbed...
Name: GarConst.xls
Type: application/msexcel
Size: 17404 bytes
Desc: not available
URL: <>

More information about the SIGMETRICS mailing list