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

[David Goodman]

An earlier posting is correct, that minor
differences in ranking based on IF are statistically
meaningless, especially near the bottom. This is related
to the choice of non-parametric statistic, and everything
I know about that I learned from Bensman's papers; in this
context http://www.garfield.library.upenn.edu/bensman/jasistprobart.pdf
. When examing a ranked list in a subject,
 the safest way is quartiles; some publishers I
am told use 1/ 3 ,  2/3 ; I usually use top 1/10 and top 1/2.
This to a certain extent gives a way of comparison  between
different subjects. One must still bear in mind that
in some subjects the really minor journals are included, and in
other they are not.
In biological subjects at least,
I think this correponds to actual human empircal decisions--
the top tenth usually corresopond to the general recognition
of the essential journals;  the top half is the titles
which are generally subscribed to by all major libraries;
the bottom half is where only libraries with a special interest
would get. And remember there is a substraum of English
language Western-country journals which are not even
included, and a penubra of non-English language journals
or English non-Western countries's journals which ought to be in,
and would  increase the validity of  JCR -based work.

A useful device when dealing with journals published irregularly,
which can be seen readily enough by the variation in cited article counts
is to take a moving average (I find 2-years sufficient.)
If one wishes to avoid zeros in one's equation, a convenient practice
is to transform all the values by add +1 (article or citation, as needed)
If you read the online information in how title changes are
handled, it essentially comes down to  the inability to use  the data for
the following two years and the preceeding two years, unless you
add the values up yourself. Before you can do this safely you must look at the
actual journals; fortunately , this can be done online regardless of access, ,
because the tables of contents are generally enough.

One of the running threads of this discussion seems to have been
the practice of  using instruments that one does not fully understand.
I will say a little about Ulrich's next week.
(One could also add the practice of librarians with insufficient statisical
knowledge to select and make proper use of the right instrument. But
librarians generally have enough sense to know to work together with a statistician
when this becomes non-trivial, wheras a few  information scientist seem
to assume they know everything pertaining to information, with all
other professions in a lesser role..)

Dr. David Goodman
Associate Professor
Palmer School of Library and Information Science
Long Island University
and formerly
Princeton University Library

dgoodman at liu.edu
dgoodman at princeton.edu
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.asis.org/pipermail/sigmetrics/attachments/20060310/c1f2f30e/attachment.html>