Examples for counterintuitive behavior of the new citation-rank indicator P100 for bibliometric evaluations

TL;DR

This paper examines the P100 citation-impact indicator across various datasets, revealing counterintuitive behaviors and limitations in its application for bibliometric evaluations, especially over time and across fields.

Contribution

It provides a detailed analysis of P100's behavior on multiple datasets, highlighting its potential issues and influencing factors in bibliometric assessments.

Findings

P100 shows counterintuitive results in model datasets

Performance can vary with journal selection

Similar issues observed in empirical datasets

Abstract

A new percentile-based rating scale P100 has recently been proposed to describe the citation impact in terms of the distribution of the unique citation values. Here I investigate P100 for 5 example datasets, two simple fictitious models and three larger empirical samples. Counterintuitive behavior is demonstrated in the model datasets, pointing to difficulties when the evolution with time of the indicator is analyzed or when different fields or publication years are compared. It is shown that similar problems can occur for the three larger datasets of empirical citation values. Further, it is observed that the performance evalution result in terms of percentiles can be influenced by selecting different journals for publication of a manuscript.