Comparing journal impact factors across subject fields?

[Loet Leydesdorff]

At 10:44 20-3-2003 -0500, you wrote:
>I am doing a study in which I would like to compare the journal impact
factors and rankings across subject fields (as given by the Journal
Citation Reports). Given the skewed nature of the data and the differences
in the citation habits of each subject field, I've been searching for a way
to do this so that I am still comparing apples to apples (i.e., like
results to like results), perhaps similar to z-scores for normal
distributions. I've been looking in the literature for such a means but
without much luck. Can anyone suggest an article (or two) that would
explain such a method (if it exists) and demonstrate how it is used?
>
>Many thanks for your help in this matter.
>
>Regards,
>

Indeed, you have two problems here:

1. the skewed nature of the underlying data
2. the grouping into disciplinary structures

ad 1.
The impact factor is based on the means and not on the mode. Peter van den
Besselaar suggested a method that can be applied if the groups to be
compared are already specified. In that case, one can perhaps also work
with the full distribution using entropy statistics. The Theil index (for
inequality) can perhaps provide a starting point for the elaboration.

ad 2.
Ranking and grouping are two very different operations. Ranking is a
consequence of hierarchical ordering of relations between journals, while
grouping is defined at the level of the resulting network among journals
and in terms of positions. Within each group one can then again make a
rank. Between groups one expects interfaces. Sometimes these interfaces
have a higher status in the hierarchy.

For example, the journal _Limnology and Oceanograpy_ provided an interface
between hydrobiology and marine biology (in the years that I looked at it),
and it also had a higher impact factor. In other cases, the
"interdisciplinary" journals are more distributed and sometimes marginal.

A way forward might be to work with groups of journals as macro-journals.
As a relatively easy method one could make the matrix of aggregated
journal-journal citations of the group of journals that one is interested
in and submit this matrix to hierarchical clustering along the cited
dimension. This would provide you with a dendogram as a representation of
the hierarchy in the data.

With kind regards,


Loet

---------------------------------------------------------
Loet Leydesdorff
Amsterdam School of Communications Research (ASCoR)
Kloveniersburgwal 48, 1012 CX  Amsterdam, The Netherlands
Tel.: +31-20-525 6598; fax: +31-20-525 3681
loet at leydesdorff.net ; http://www.leydesdorff.net
http://www.upublish.com/books/leydesdorff.htm