Graphs used in Essential Science Indicators^{SM -} Projecting Full-Year Citation Counts.

Essential Science Indicatorsplots time series graphs for scientists, institutions, journals and countries. These graphs can show trends in citations received, papers published, and citation impact, that is, average citations per paper. The time series consist of a series of overlapping five-year periods, e.g., 1993-1997, through the most recent period 1999-2003. Since Essential Science Indicators is updated every two months, in bimonthly increments, the most recent year, say 2003, may contain less than a full year of citations and papers that can make the last time period appear abnormally low. For example, after the first bimonthly update of 2003, the last five-year time period would consist of 4 years and 2 months of data; after the next update, 4 years and 4 months, and so on. The result is that, without correction, all time series graphs occurring during a mid-year cycle will show smaller values for the last period than are likely to exist by year-end.

To correct for this, Essential Science Indicators estimates full-year scores for citations, papers and cites per paper for scientists, institutions, journals and countries. The process involves multiplying the actual values for the last five-year period in the time series by the appropriate factor from the table below.

For example, if the Essential Science Indicators bimonthly update period is 2, and the counts for citations, papers and cites/paper are 20, 10 and 2.0 respectively, the full-year estimates would be 20x1.37 = 27.4 citations, 10x1.15 = 11.5 papers, and 2.0x1.19 = 2.4 cites per paper. After the last (6^th) bimonthly update of the year, the multipliers revert to 1 indicating that no estimation is needed since the year is complete.

The derivation of multipliers is based on the empirical observation that, across the entire database, the last year of a five-year citation window contributes on average 41% of the citations to the total for the period. The further approximation is made that the 41% of citations are spread evenly across the six bimonthly periods, although in a more sophisticated scheme, the later periods would contribute more than the earlier ones due to the widening citation window. It is also assumed is that the publication rate can be projected evenly, that is, each year will contribute the same number of papers. Thus, the estimate does not take into account whether an entity has an increasing or decreasing rate of publication or citation, or possible field differences in growth rates.

Should you have further questions, please contact us.