How to improve the prediction based on citation impact percentiles for years shortly after the publication date?

• Is it possible to improve citation impact measurements by the consideration of journal impact and other covariates?
• The dataset contains 475,391 journal papers published in 1980 and their annual citation counts (from 1980 to 2010).
• We use percentiles of citations as an indicator of citation impact.
• Our results show that citation impact measurement shortly after publication can really be improved by considering covariates.

The findings of Bornmann, Leydesdorff, and Wang (2013b) revealed that the consideration of journal impact improves the prediction of long-term citation impact. This paper further explores the possibility of improving citation impact measurements on the base of a short citation window by the consideration of journal impact and other variables, such as the number of authors, the number of cited references, and the number of pages. The dataset contains 475,391 journal papers published in 1980 and indexed in Web of Science (WoS, Thomson Reuters), and all annual citation counts (from 1980 to 2010) for these papers. As an indicator of citation impact, we used percentiles of citations calculated using the approach of Hazen (1914). Our results show that citation impact measurement can really be improved: If factors generally influencing citation impact are considered in the statistical analysis, the explained variance in the long-term citation impact can be much increased. However, this increase is only visible when using the years shortly after publication but not when using later years.