The generalized pair-correlation method (GPCM) is a nonparametric, pairwise ranking method originally designed for variable selection in chemometrics [1]. In GPCM, variables are compared in all possible pairs by arranging outcomes into 2×2 contingency tables and applying statistical tests to identify a winner, a loser, or no decision; the final ranking can be based on simple wins, win–loss differences, or probability-weighted differences. Because this framework is robust to non-normality, collinearity, and mixed strength of associations, it is attractive well beyond classical descriptor selection. A promising new application is the regeneration and investigation of Mendeleev’s periodic table from measured or computed elemental properties. The periodic table is fundamentally an ordered system of 118 elements arranged by atomic number, with recurring similarity patterns across groups and periods; historically, Mendeleev’s key achievement was to organize known elements by recurring chemical behaviour and to predict missing ones from gaps in that structure [2]. Modern formal studies likewise describe the periodic system as a structure built from relations of order and similarity [3] rather than from a single property alone. In this context, GPCM can be used as a chemometric engine for reconstructing periodic regularities from multivariate elemental data. Instead of forcing the properties into one global linear model, GPCM can compare them pairwise with respect to a target criterion such as group similarity, period continuity, metallicity, or a composite periodicity score, and then rank which properties best recover known periodic structure [3]. This approach offers several research advantages:
- First, GPCM can identify which elemental properties are the most informative for regenerating the accepted periodic arrangement and which are redundant or weakly discriminating.
- Second, its natural tendency to form tied groups may be useful rather than problematic, because degeneracy can reveal families of elements or descriptors that are genuinely similar, echoing the similarity classes that underlie the periodic law.
- Third, by repeating the analysis with incomplete historical datasets, one could simulate a “Mendeleev-like” discovery setting and test which subsets of properties are sufficient to recover missing-element patterns or plausible placements for anomalous elements.
New developments could extend GPCM from simple ranking to periodic-system discovery workflows. Coupling GPCM with modern correlation analysis, clustering, seriation, and validation methods would allow researchers to move from pairwise comparisons to interpretable maps of elemental similarity and periodic neighbourhoods [4]. In this role, GPCM would not replace quantum theory or atomic-number ordering, but it could provide a transparent, data-driven bridge between historical classification, modern chemometrics, and explainable materials informatics.