Cause of NLCS’s weaker performance relative to NCS: errors or fundamental factors

Ascertain whether the observed weaker correlation of the Normalised Log-transformed Citation Score (NLCS) compared to the Normalised Citation Score (NCS) when using Scopus and OpenAlex data is attributable to avoidable document classification or metadata errors (e.g., inclusion of editorials or uncitable items), or to fundamental characteristics of the scholarly record and citation distributions (e.g., genuine prevalence of uncitable or short-form articles), thereby determining if NLCS’s limitations can be mitigated in principle.

Background

NLCS was designed to address skewed citation distributions by log-transforming counts before normalisation, intending to reduce undue influence of highly cited papers on field averages. Empirically in this paper, NCS often performed better than NLCS across fields, raising questions about the source of NLCS’s relatively weaker results.

The authors suggest potential causes, including classification and document-type errors (e.g., editorials treated as articles) and structural properties of the literature (e.g., many legitimate but rarely cited short-form pieces). Clarifying whether the issue is error-driven or inherent is necessary to understand if NLCS can be improved or if its limitations are fundamental.