"Informationswissenschaftliche Zeitschriften in szientometrischer Analyse"

[Eugene Garfield]

Title     Informationswissenschaftliche Zeitschriften in
          szientometrischer Analyse
Fachhochschule Koeln
Grazia, Colonia, July 2002


Full Text in German available at :
http://www.fbi.fh-koeln.de/fachbereich/papers/kabi/volltexte/band033.pdf


Translation of paragraph 5.3:

Garfield's conjecture: Do international journals have a higher impact factor
than German-language journals?

Eugene Garfield postulates: German-language science journals have definitely
their regional significance, but cannot compete well with internationally
oriented periodicals regarding the impact factor.  He suggests that German
journals do not meet international standards, and do not report the results
of top research.., since even German scientists prefer to offer their best
work to international journals: ''I am often surprised at the level of
misunderstanding of German editors about citation analysis and the impact
factor. There is a great need for national journals written in German and
other European languages, but it is absurd to expect them to reach the same
level of impact as international journals, unless they adhere to the same
standards as leading journals and attract significant original research.
German scientists understand that fact of life and for this reason they
mainly publish their best work in English in international journals''
(Garfield 2002, 25).  Does this conjecture also hold true for
German-language information science?

A cursory look at Table 5-3 shows that German-language journals are
liberally.... distributed in the classification of international
periodicals.  In no way do they rank last according the impact factor.
Lediglich, purely, simply, ZfBB and NfD, however, are represented in the ISI
databases.  As for «Bibliothek. Forschung und Praxis», «ABI Technik»,
«Bibliotheksdienst, Buch und Bibliothek» and «ProLibris», their exclusion
cannot then be explained.  Hence, the omission of this titles must coincide
with some unmet ''basic standards'' (cf. Chapter 1).  However, some of our
periodicals do meet those standards.  The journal publication dates are
regular; the articles have well-formulated searchable titles; footnotes are
available (otherwise, we would have not been able to count them); English
summaries are provided.  If the texts are always peer-reviewed, that we
cannot verify.  (''JfD'' has no peer review process, but stills shows up in
ISI's products.) It appears as though ISI is covering two German alibi
journals with ''ZfBB'' and ''NfD'' without involving itself any further in
the area of German periodicals.  It is also possible that ISI does not know
these periodicals at all, because they are hardly cited in ISI source
journals.  This could just be evidence then of poor marketing on the part of
publishers and publishing companies.  In short, ISI is thoroughly open to
expert opinions -- and that means in plain talk ''knocking on doors''!  So
certainly for some of the top German journals in information science,
Garfield's conjecture does not apply.

However, Eugene Garfield may not have interpreted his conjecture in this
fashion. Still this matters involves, to a greater extent, German-language
and international journals in general.  After this preliminary glance, we
must look more closely and compare both distributions (German-language
periodicals based on the rIF and JCR periodicals based on the IF) using
statistical methods.  We consider the 38 volumes of the German periodicals
and the 144 volumes of the international periodicals as sampled sets.  (The
samples are not random, as we have explained in the methodology section.
Therefore, the results should be read with some reservations.)  We determine
the confidence interval which contains the true value of the population
parameter.  If the intervals do not intersect, then the distributions are
different and Garfield's conjecture would be confirmed.  If the intervals do
in fact intersect, then the distributions are not different and Garfield's
conjecture would be quite dubious.  We accept a 1%-error probability so that
our result will have a 99%-significance level.

For the German information science journals, we find:
- arithmetic mean: rIF = 0.25
- standard deviation of the sample: 0.19
- N = 38
- Confidence level (1%): 0.08
-The value of the mean for German periodicals lies between 0.17 and 0.33 for
a 1%-percent error rate.

The result is clear.  Both intervals have different ranges so that we may
assume different parameters.  The average German information-science journal
(statistically speaking) has a regional impact factor lying presumably in
the range 0.17-0.33.  The average international (JCR) periodical has an
impact factor lying in the range 0.38-0.56. Therefore, the impact factor for
JCR journals is double that of the German-language periodicals.  Garfield's
conjecture is confirmed.  German information science journals have a
significantly lower impact factor than the international periodicals
evaluated by ISI.