On the relationships between bibliographic characteristics of scientific documents and citation and Mendeley readership counts: A large-scale analysis of Web of Science publications

On the relationships between bibliographic characteristics of scientific documents and citation and Mendeley readership counts: A large-scale analysis of Web of Science publications1

Zohreh Zahedi* & Stefanie Haustein**

• z.zahedi.2@cwts.leidenuniv.nl *Centre for Science and Technology Studies (CWTS), Leiden University, Wassenaarseweg 62A, Leiden, 2333 AL (The Netherlands)

**stefanie.haustein@umontreal.ca **École de bibliothéconomie et des sciences de l’information, Université de Montréal, C.P. 6128, Succ. Centre-Ville, Montréal, H3C 3J7 (Canada)   Abstract In this paper we present a first large-scale analysis of the relationship between Mendeley readership and citation counts with particular documents’ bibliographic characteristics. A data set of 1.3 million publications from different fields published in journals covered by the Web of Science (WoS) has been analyzed. This work reveals that document types that are often excluded from citation analysis due to their lower citation values, like editorial materials, letters, or news items, are strongly covered and saved in Mendeley, suggesting that Mendeley readership can reliably inform the analysis of these document types. Findings show that collaborative papers are frequently saved in Mendeley, which is similar to what is observed for citations. The relationship between readership and the length of titles and number of pages, however, is weaker than for the same relationship observed for citations. The analysis of different disciplines also points to different patterns in the relationship between several document characteristics, readership, and citation counts. Overall, results highlight that although disciplinary differences exist, readership counts are related to similar bibliographic characteristics as those related to citation counts, reinforcing the idea that Mendeley readership and citations capture a similar concept of impact, although they cannot be considered as equivalent indicators.

Keywords Mendeley readership; WoS Citation; Bibliographic characteristics; Document types

                                                            1 This is a preprint of the paper accepted for publication in Journal of Informetrics, DOI: 10.1016/j.joi.2017.12.005  2    Introduction

Effect of document characteristics on citation impact Measuring research impact using citation analysis has a long tradition in the field of scientometrics. Today, citation-based indicators are widely used and play a central role in the evaluation of scientific works. Despite their de facto use as proxies of scientific quality, citations are not able to fully capture the use and influence of scientific papers (Moed, 2005; MacRoberts & MacRoberts, 2017). Bibliometric research has also shown that a variety of factors can influence citation counts (Opthof & Leydesdorff, 2010; Waltman et al., 2011; Larivière & Gingras, 2011). Such factors include, the document types and age of publications, their number of pages, the length of their titles and reference lists (Bornmann & Leydesdorff, 2015; Bornmann, Leydesdorff, & Wang, 2014; Vieira & Gomes, 2010); their different theoretical or methodological approaches (Antonakis, et al., 2014); whether they are open access (Hajjem, Harnad, & Gingras, 2006); the citation propensity of their fields and their interdisciplinarity (Yegros-Yegros, Rafols, & D’Este, 2015); or the Impact Factor of their publication journal (Boyack & Klavans, 2005).

Numerous previous studies have analyzed whether citation impact is affected by various document characteristics. These studies have explored different characteristics at the article, journal, and author levels using correlation and regression analyses. For example, in the Natural, Life, and Health sciences (Thelwall, 2017), papers with unusual and obscure titles were associated with lower citation impact. Mixed results were found regarding the effect of title length (Stremersch, et al., 2015; Jacques & Sebire, 2010), or titles that included non- alphanumeric characters such as hyphens or colons (Buter & Van Raan, 2011; Haslam, et al., 2008; Nair & Gibbert, 2016). Based on the assumption that longer articles with longer reference lists may reflect in-depth analysis and diversity of ideas, the number of pages and references have also been analyzed as factors that may affect citation counts (Fox & Boris, 2016). The results showed that papers with more references and more pages tended to get more citations (Ajiferuke & Famoye, 2015; Davis, et al., 2001). Similarly, the number of authors, institutes, and countries involved in a given publication may indicate the extent of collaboration, which is again assumed to increase citation impact. However, results regarding the effect of collaboration on citation rates are mixed (for an overview see Onodera & Yoshikane, 2015) as regards variations by country of collaboration (Thelwall & Sud, 2016)

…(Full text truncated)…