Braun Tibor, Glänzel Wolfgang, Schubert András: Országok, szakterületek, folyóiratok tudománymetriai mutatószámai 1981-1985 (A MTAK Informatikai És Tudományelemzési Sorozata 6., 1992)
Methodology
A. SCHUBERT, W. GLÄNZEL, T. BRAUN : SCIENTUMETRIC DATAFILES The same journal might be classified into at most three subfields. Because of the subfield overlaps, subfield counts are, in general, not additive. Major field counts were calculated by filtering out duplicates. Ad 5. The filing year of a paper may be chosen according to the date indicated on the cover of the journal {publication year) or the date of its recording into the database (tape year). It should be noted that the true publication date is somewhere in between. Publication year data are apparently more "database independent" and reproducible, therefore, this was our choice in this study. We must, however, remember that the last year's (1985) counts are somewhat deflated because of the unavoidable delays of incorporating journals issued in the last months of the year into the database. We tried to balance this by using percentage world shares rather than absolute counts in the time series charts. Even though, national differences in database delays (due to different mailing speed etc.) may cause undesirable biasses. Ad 6. In some of our earlier papers 1,1 2, source and citation periods used in the Journal Citation Reports 1 1 were imitated: source items published in a twoyear period and citations to them in the subsequent (third) year were counted. Some critics objected that this distance of three years between cited and citing documents is not enough to gain a realistic picture of citation impact. (Indeed, citations peak usually 2-3 years after publication. A half-year delay in this period may, therefore, cause dramatic changes in citation rate.) On the other hand, choosing too long periods make part of the data outdated and long-range changes (e.g. journal demography) non-negligible. In this study the same five-year interval (1981-1985) is used both as source and citation period. This means that the actual citation period varies from 0 to 5 years. About half of the papers under study reach their citation peak within the period. On the other hand, as mentioned above, database changes have no dramatic effect within this period: papers published in the fix journal set are responsible for 98% of the citations to all papers covered by the SCI. Ad 7. For searching citations to source items, an identification key of each item of the source file has been built. This key is a 15 character string made up from the letters of the author's name, the journal title and the digits of the publication year, volume and page. The same identification keys are then formed of the cited item side of the citation files. Citations are then counted as matches on the two files. 8 Scicntomctrics 16 (19R9)
