Source-linked AI summary
On the robustness of the h-index
Jerome K Vanclay
TL;DR
The paper examines limitations of the journal impact factor and evaluates the h-index’s robustness and verifiability. It finds that the h-index avoids long-tail citation problems and supports favouring a Hirsch-type index.
Problem
The journal impact factor has recognised flaws, motivating assessment of alternatives such as the h-index.
Method
The paper assesses the h-index’s robustness to citation-distribution errors and the practicality of verifying it.
Results
The h-index avoids many long-tail citation problems and is relatively unaffected by citation errors.
Takeaways & Limitations
Robustness and ease of verification support favouring a Hirsch-type index over the journal impact factor.
Takeaways & Limitations
Error-free verification can be difficult without intimate knowledge of the candidate publications.
Abstract
from arXiv · showhide
The h-index (Hirsch, 2005) is robust, remaining relatively unaffected by errors in the long tails of the citations-rank distribution, such as typographic errors that short-change frequently-cited papers and create bogus additional records. This robustness, and the ease with which h-indices can be verified, support the use of a Hirsch-type index over alternatives such as the journal impact factor. These merits of the h-index apply to both individuals and to journals.
Introduction
The introduction presents the h-index as an alternative to the flawed and error-prone journal impact factor. It emphasizes the h-index’s robustness to long-tail distortions and its ease of verification for individuals and journals.
- Motivation: The journal impact factor is influenced by recognized flaws, while the h-index avoids several problems including censorship, errors, manipulation, and long-tailed distributions.The h-index was initially proposed for individual scientists, with extensions suggested for teams and journals.
- Problems with impact factors: Censorship, manipulation, and selection effects make citation and publication counts approximate and often biased, producing considerable error in inferred impact factors.These problems are likely greatest in both tails of the citation distribution.
- Robustness: The h-index avoids many such issues by ignoring distributional long tails and focusing on the middle part of the Zipf plot.This makes it relatively unaffected by errors concentrated in citation-distribution tails.
- Verification: As an integer, the h-index avoids the false precision of three-decimal impact factors and is easier to verify than indices based on mean citations.Disputes over total citation counts or mean-based indices can be difficult to resolve reliably.
- Verification: Verification usually requires checking only a few publications near the threshold, rather than the total citation count.Checks determine whether typographic errors concealed one or two citations that could raise an index from n to n+1.
Approach and Methods
The h-index was evaluated using citation records from Google Scholar and Web of Science, with obvious errors corrected and selected records cross-checked. Across individual and journal examples, the index remained stable despite discrepancies, supporting its robustness and relatively simple verification.
- Verification procedure: Corrections generally mattered only for publications ranked below the preliminary h-index, so verification focused on threshold records where combined counts could fall below rank.The citation union needed examination only for a few cases; publications already exceeding the estimated h-index could not affect it.
- Individual publication record: Despite a relatively large change in citations to one paper, the individual’s h-index did not change, and many co-author citations altered it only slightly to 13.The analysis concluded that the h-index was a robust indicator of published output in this instance, although verifying robustness for other researchers is more difficult.
- Journal application: For Forest Ecology and Management, correcting records reduced the Google Scholar and Web of Science counts to 185 and 195, respectively, without changing the h-indices.Several records changed rank in both datasets, yet the journal h-index remained unchanged.
- Journal application: Combining the larger citation count from the two databases produced a journal h-index of 29, which remained stable across the diverse sources despite many raw-data discrepancies.The discrepancy between the two naïve h-indices was about 15%.
Conclusion
The h-index’s integer nature, robustness to perturbations in the tails of the publication-citations distribution, and ease of verification support favouring a Hirsch-type index over one based on total citations and publications.
- The h-index is integer-valued.
- It remains robust to perturbations in the tails of the publication-citations distribution.
- Its ease of verification offers a practical advantage.
- These properties support favouring a Hirsch-type index over an index based on total citations and total publications.