The Rn-index: a more accurate variant of the Rk-index

Alonso Rodriguez-Navarro

REVIEW 3 major objections 5 minor 4 references

The Rn-index: a more accurate variant of the Rk-index

T0 review · 3 major / 5 minor · reviewed 2026-08-12 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read The Rn-index, a simple sum of local-to-global rank ratios, corrects the Rk-index's saturation and summability failures for countries that dominate research.

desk verdict A simple arithmetic-mean variant of the Rk-index that fixes saturation and summability issues in high-output countries, but the accuracy claim rests on a validation transfer that doesn't cover the regime it's meant to fix. read the letter →

arxiv 2411.18161 v1 pith:XMARY6OY submitted 2024-11-27 cs.DL cs.IR

Alonso Rodriguez-Navarro This is my paper · ORCID

classification cs.DLcs.IR

keywords Rn-indexRk-indexrankratiosresearchperformanceevaluationbreakthroughpaperssummabilitycitationanalysistopcited

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper proposes the Rn-index, a variant of the Rk-index, to measure a country's or institution's contribution to pushing the boundaries of knowledge, a contribution that ordinary citation indicators often miss because breakthrough papers are only about 0.01% of all publications. The Rk-index works well in general but saturates when a few actors produce most advances: for the USA in lithium batteries, its value sits near the index's maximum, so differences between leading countries are compressed and the index is not summable when domestic and collaborative papers are pooled. The Rn-index replaces the Rk calculation with an arithmetic sum of ten local-to-global rank ratios, which spreads out the compressed values and makes summability deviations smaller and independent of the index's magnitude, as shown for the USA, China, and other countries across topics such as solar cells, lithium batteries, and stem cells. A reader should care because research performance at the frontier is used to compare countries and institutions, and a simple, fractionable indicator that avoids this particular distortion would make such comparisons more trustworthy.

What carries the argument

The carrying object is the Rn-index, defined as ten times the sum, over a country's or institution's ten most-cited papers, of the ratio of each paper's local rank to its global rank. The Rk-index is the geometric mean of inverted global ranks of the ten most-cited papers with 20 added, and the Rn-index replaces that geometric mean with an arithmetic mean of rank ratios. This arithmetic mean is what gives the indicator its corrective behavior: it remains sensitive to positions inside the very top of the global distribution, where Rk values pile up against their maximum, and it treats each of the ten papers additively, which is what makes the indicator closer to summable when domestic and collaborative sets are pooled. The paper uses the observed strong relationship between Rn and Rk, and between arithmetic and geometric means of rank ratios, to argue that the established synthetic-series validation of Rk carries over to Rn.

What would settle it

Take a set of synthetic citation series with known breakthrough papers at the 0.01% level, compute Rn and Rk for many synthetic countries, and check whether Rn reproduces the known ordering and has smaller summability deviations than Rk in high-output cases; alternatively, compute Rn and true top-0.01% counts in a large multi-topic dataset and see whether the reported correlation with the 0.1% proxy survives.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that computing ten times the sum of the ratios between local and global ranks of a country's ten most-cited papers yields an indicator, the Rn-index, that behaves better than the Rk-index exactly where the Rk-index is weakest. For high-output countries the Rn-index separates domestic and collaborative publication sets that the Rk-index leaves nearly indistinguishable, and it reduces or eliminates the inflation of values that occurs when the Rk-index of domestic papers is added to that of collaborative papers. The paper reports that in the USA lithium-battery case the Rk sum exceeds the pooled Rk by 48.9%, while the Rn sum deviates by only -2.0%; across the tested countries and topics the Rn deviations are smaller and depend only on randomness, not on the size of the indicator. The paper also reports a correlation of 0.65 (p = 0.001) between Rn values and counts of top 0.1% cited papers, which it reads as external support for the indicator while noting that percentile counts discard the within-percentile information that Rn preserves.

Load-bearing premise

The load-bearing premise is that the Rk-index's synthetic-series validation transfers to the Rn-index because the two are numerically close, and that counts of top-0.1% cited papers are an acceptable stand-in for the true top-0.01% breakthrough rate; the paper's Section 4 explicitly says "Assuming this" for the second step, so if either link fails the claim that Rn is more accurate is not established.

Editorial extensions

If this is right

For countries like the USA and China in fields such as lithium batteries and solar cells, Rn-based country comparisons will show larger and more meaningful gaps than Rk-based comparisons.
Rn is closer to satisfying the summability property, so combining domestic and collaborative papers, or institutions, will not inflate the total as much as Rk does.
Because Rn is just a sum of ten rank ratios, it is easy to compute, explain, and use with fractional counting for co-authored papers.
If the correlation with top-0.1% cited papers holds more broadly, Rn can serve as a lightweight indicator of breakthrough-level contribution before enough top-0.01% papers accumulate to measure directly.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

One could test whether Rn's summability advantage grows with topic concentration: the more skewed the world's output toward a few countries, the more Rn should outperform Rk.
Because the ratio of local to global rank has a bounded range, Rn may admit analytic distributional results, such as moments or asymptotic bounds, that Rk's extreme-order statistics do not, which could make significance tests for country differences straightforward.
If Rn is adopted for country-level frontier assessment, its sensitivity to the choice of the ten-paper window should be probed; the paper does not analyze what happens when the window size changes.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

The Rn-index is a simple arithmetic-mean variant of the Rk-index, and the paper gives a fair, transparent first look at why it might be preferable for dominant countries. The idea is new relative to the references, easy to compute, and fractionally countable, which are real practical advantages. The summability data in Table 3 are the strongest part: across the 12 country-topic pairs, Rn deviations are smaller than Rk's and do not grow with the value of the index, fixing the saturation pattern that motivated the paper. That is worth credit.

The soft spot is the validation transfer in Section 4. The paper concedes in Section 2 that the Rk-index's weakness appears precisely because the saturation regime "was not tested with synthetic series." Then Section 4 transfers Rk's synthetic-series validation to Rn because of the strong Rk-Rn relationship in Fig. 1. But that relationship is upward-curving at high Rk values, so the transfer is an extrapolation into the very regime where Rk's own validity is unknown. Table 4 does not close the gap: it correlates Rn with top 0.1% cited papers (r=0.65, no CI) and includes no Rk baseline on the same 23 units, so we cannot tell whether Rn is actually more accurate than Rk at the top 0.01% frontier. The author explicitly flags the 0.1% assumption, which is good, but the claim that Rn is "more accurate" is not yet supported.

These are fixable issues. A head-to-head Rk vs Rn regression on the same benchmark and a sensitivity check at 0.01% would go a long way. The sample of Table 3 is small and non-random, so treat the summability improvement as suggestive rather than proven. In proportion, the paper is a coherent, honest proposal of a plausible indicator, not an overclaiming one; the limitations are largely admitted.

I'd send this to peer review. The index is simple enough to be used, and referees can settle the validation question with a modest amount of additional analysis. I wouldn't cite it in my own work this year, but I'd bring it to a reading group interested in bibliometric indicators.

Referee Report

3 major / 5 minor

Summary. The paper introduces the Rn-index, a variant of the Rk-index designed to correct what the author identifies as a weakness of the Rk-index: saturation and loss of summability when a small number of countries produce most of the highly cited papers in a topic. The Rn-index is defined as the sum of the ratios between local and global ranks of the 10 most cited papers, multiplied by 10 (equivalently, the arithmetic mean of these ratios scaled by 100). Using data for several countries and topics, the paper shows that Rn-values deviate less from the summability property than Rk-values and spread out values that saturate near the Rk maximum. The paper then argues that the synthetic-series validation of the Rk-index transfers to the Rn-index because of a strong numerical relationship between the two indicators, and it reports a real-data correlation of r = 0.65 (p = 0.001) between the Rn-index and the number of top 0.1% cited papers across 23 country–topic units. The author concludes that the Rn-index provides a more accurate measure of contribution to scientific advancement than the Rk-index, especially for high-output countries.

Significance. If the central claim is correct, the Rn-index would be a practically useful, easily computable bibliometric indicator that avoids a known saturation artifact of the Rk-index while preserving its intended interpretation. The paper is honest about the rarity of breakthrough-level papers and about the proxy nature of the top 0.1% threshold. The summability improvements in Table 3 are clearly presented and, within the small sample, support the claim that the Rn-index behaves better than the Rk-index on that specific property. The paper also has the virtue of transparency: the index is defined by an explicit formula with no fitted parameters beyond the k = 10 and offset +20 inherited from the Rk-index. However, the significance of the paper for the journal rests on demonstrating that the Rn-index is more accurate for measuring frontier contribution, and that demonstration currently depends on an extrapolation of prior validation results into a regime that the original validation did not cover, plus a moderate proxy correlation without a head-to-head comparison against the Rk-index.

major comments (3)

[Section 4, first paragraph] The transfer of the synthetic-series validation from the Rk-index to the Rn-index is not valid for the regime in question. Section 2 explicitly states that the Rk-index's weakness arises at values close to its maximum and that this situation 'was not tested with synthetic series.' Section 4 then transfers the validation 'considering the strong numerical relationship between the Rn and Rk indices (Fig. 1).' But Fig. 1 shows upward curvature at high Rk values, so the strong relationship does not imply that the Rn-index inherits Rk's validated behavior precisely in the high-value regime where Rk was never validated. The paper needs to either provide synthetic-series tests that include the saturation regime or otherwise justify the transfer for that regime.
[Section 4, Table 4] The real-data validation in Table 4 does not close the gap in the accuracy claim because (i) the relevant breakthrough threshold is top 0.01% cited papers, as the paper itself states, but the analysis uses top 0.1% with the phrase 'Assuming this,' without any evidence that the coarser percentile behaves similarly for the countries and topics studied; and (ii) no Rk-index is computed for the same 23 country–topic units, so the table cannot show that the Rn-index is more accurate than the Rk-index, only that the Rn-index correlates moderately with a top-percentile count. A head-to-head comparison of Rn and Rk against the same benchmark is needed to support the comparative claim.
[Table 3 and surrounding text] The summability comparison in Table 3 is based on a small, non-random set of countries and topics (12 rows) and is not accompanied by any statistical test. The text states that Rk deviations are 'random, but also larger for higher values' and that Rn deviations are 'only random, without any dependency on high values,' but these claims are based on visual inspection without a quantitative test of the deviation–magnitude relationship. A simple regression or correlation between deviation size and index magnitude, with confidence intervals, would substantiate the claim that the Rn-index removes the value-dependence of the deviation.

minor comments (5)

[Section 4 and Table 4] The paper reports Pearson's r = 0.65 and a two-sided p-value but does not report a confidence interval or the effective sample size per topic; adding these would clarify the strength of the correlation given the small number of units.
[Table 3] The abbreviation 'Doma' appears without a footnote in the table header; the footnote defining 'Dom, domestic' is present but would be easier to use if placed immediately under the table.
[References] There are typos in two references: 'reserach' should be 'research' in the 2024a and 2024b entries, and 'knowlege' should be 'knowledge' in the 2024c entry.
[Section 3, last paragraph] The sentence 'Furthermore, in these tests, the deviations of the Rn-index are smaller than those of the Rk-index' is ambiguous because Table 3 shows that in some rows (e.g., China lithium batteries) the Rn deviation is not smaller than the Rk deviation; the generalization should be qualified.
[Section 3, Table 2] The definition of the Rn-index as 'sum of the 10 rank ratios multiplied by 10' is not immediately clear from the table, which shows the arithmetic mean of rank ratios; a sentence clarifying the scaling factor would avoid confusion.

Circularity Check

2 steps flagged · score 3.0 of 10

Partial circularity: Rn's validation is bridged by the author's own Rk work, and its claimed correction is largely the criterion used to select the arithmetic-mean variant.

self citation load bearing [Section 4, Validation of the Rn-index, first paragraph]
"Considering the strong numerical relationship between the Rn and Rk indices (Fig. 1), the validation of the latter with synthetic series (Rodríguez-Navarro & Brito, 2024) extends to the Rn-index."

The central accuracy claim is bridged by a self-citation: Rn inherits Rk's synthetic-series validation because the two are numerically close. But §2 states that the weakness Rn is designed to fix arises in a saturation regime that 'was not tested with synthetic series.' The cited prior validation therefore does not cover exactly the region where Rn claims superiority. The bridge (Fig. 1) is an empirical correlation in the same real-world data, not an independent check of Rn, so the validation transfer reduces to an in-sample relationship plus an unverified-for-this-regime self-citation.
other [Section 3, The Rn-index corrects the weakness of the Rk-index, paragraphs 2-3]
"Based on this finding, the arithmetic mean seemed the most appropriate indicator because it is equivalent to using the sum of the ratios... The obvious upward curvature of the Rk-Rn-index plots... along with the results shown in Table 2, indicates that the Rn-index corrects the above-described weakness of the Rk-index at high Rk-index values."

The arithmetic-mean variant is chosen after observing that the arithmetic and geometric means 'correct these weaknesses of the Rk-index' (saturation and summability deviations). The same criterion and partly the same data (Table 2, and Table 3 cases such as USA/China lithium and solar cells) are then presented as evidence that Rn 'corrects the weakness.' Thus the headline improvement over Rk is to a substantial degree the selection rule restated as a result; Table 4 provides external correlation for Rn alone, with no Rk baseline on the same units.

full rationale

The paper contains no parameter fitting and Rn is a simple, transparent variant, so this is not a fully circular derivation. However, the two moves above are self-referential: validation is imported from the author's own Rk work, and the correction is demonstrated on the criterion used to pick the variant. These make the claim 'more accurate than Rk' weaker than presented, but they do not reduce the whole derivation to an equation identity. Score 3 reflects partial, mild circularity.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The paper introduces no physical or ontological entities. The main commitments are design parameters inherited from Rk, a proxy threshold for breakthrough papers, and the assumption that prior Rk validation carries over to Rn. These are the points where the reader must accept something on faith rather than on demonstration.

free parameters (3)

Number of most cited papers (k=10) = 10
The index is fixed to the ten most cited papers, inherited from the Rk-index. The claimed correction and the summability results depend on this threshold, and no sensitivity analysis is provided.
Rank offset (+20) in Rk/Rn denominator = 20
The plus-20 offset comes from the Rk formula and defines the saturation scale. It is not independently justified for the Rn-index.
Scaling multiplier in Rn definition = 10
The Rn-index is defined as the sum of the ten rank ratios multiplied by ten. This scaling is arbitrary and not derived from any fitted criterion.

assumptions (4)

domain assumption Citation ranks of the most cited papers follow a power law or ordered deviations, so arithmetic and geometric means are linearly related.
Used in Section 3 and Figure 2 to justify choosing the arithmetic mean because Rk is also a geometric mean.
domain assumption Top 0.1% cited papers are a reasonable proxy for top 0.01% breakthrough papers.
Adopted in Section 4 to obtain enough data for correlation analysis, with no sensitivity analysis around the threshold.
ad hoc to paper The synthetic-series validation of the Rk-index transfers to the Rn-index because of the strong numerical relationship in Figure 1.
Stated in the first paragraph of Section 4 without re-running the synthetic validation on the Rn-index.
domain assumption Rows in the correlation analysis are treated as independent, despite multiple rows belonging to the same country or topic.
The Pearson test in Section 4 ignores the nested structure of the data in Table 4.

how reviews work

0 comments

Cite this review

Pith. "Pith review of The Rn-index: a more accurate variant of the Rk-index." pith.science (2026). https://pith.science/paper/XMARY6OY

@misc{pith2026241118161,
  author       = {Pith},
  title        = {Pith review of: The Rn-index: a more accurate variant of the Rk-index},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XMARY6OY}},
  note         = {Machine review of arXiv:2411.18161}
}

read the original abstract

The contribution to pushing the boundaries of knowledge is a critical metric for evaluating the research performance of countries and institutions, which in many cases is not revealed by common bibliometric indicators. The Rk-index was specifically designed to assess such contributions, and the Rn-index is a variant that corrects the weakness of the Rk-index, particularly in the evaluation of countries that produce a high proportion of global advancements. This is the case of the USA and China in many technological fields. Additionally, the Rn-index is simple to calculate and understand, as it involves only summing the ratios between the local and global ranks of papers, ordered by their citation count. Moreover, the Rn-index may also be fractionally counted.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

4 extracted references · 4 canonical work pages

[1]

Introduction Rodríguez-Navarro and Brito (2024) recently introduced the Rk-index, a non-parametric indicator that reveals the contribution to pushing the boundaries of knowledge—an essential metric for evaluating research in countries and institutions (Rodríguez-Navarro, 2024a). However, the calculation of this contribution is challenging because the invo...

work page 2024
[2]

Calculation of indicators based on the ranks of most cited papers: Rk-index and means of rank ratios. Assessment of China in composite materials Rank2a Collaborative publications Domestic publications All publications Rank1a 1/ (20+Rank1) Rank1/ Rank2 Rank1 1/ (20+Rank1) Rank1/ Rank2 Rank1 1/ (20+Rank1) Rank1/ Rank2 1 1 0.048 1.000 5 0.040 0.200 1 0.048 1...

work page 2011
[4]

This would necessitate the integration of many topics or the use of very wide citation windows

Correlation between the number of top 0.1% cited papers and the corresponding Rn-index in several countries and research topics Country (type of papers) Topic top 0.1% Rn-index USA (collaborative) Stem cells 39 46.6 USA (domestic) Stem cells 31 58.8 EU (collaborative) Stem cells 29 32.1 USA (collaborative) Graphene 25 50.8 China (collaborative) Graphene 2...

work page 2018
[5]

Conclusions The contribution to pushing the boundaries of knowledge is a critical metric for evaluating the research performance of countries and institutions, which in many cases is not revealed by common bibliometric indicators (Rodríguez-Navarro, 2024a; 2024b). The Rk-index was specifically designed to assess such contributions, and the Rn-index is a v...

work page Pith review arXiv 2013

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.

[1] [1]

Introduction Rodríguez-Navarro and Brito (2024) recently introduced the Rk-index, a non-parametric indicator that reveals the contribution to pushing the boundaries of knowledge—an essential metric for evaluating research in countries and institutions (Rodríguez-Navarro, 2024a). However, the calculation of this contribution is challenging because the invo...

work page 2024

[2] [2]

Calculation of indicators based on the ranks of most cited papers: Rk-index and means of rank ratios. Assessment of China in composite materials Rank2a Collaborative publications Domestic publications All publications Rank1a 1/ (20+Rank1) Rank1/ Rank2 Rank1 1/ (20+Rank1) Rank1/ Rank2 Rank1 1/ (20+Rank1) Rank1/ Rank2 1 1 0.048 1.000 5 0.040 0.200 1 0.048 1...

work page 2011

[3] [4]

This would necessitate the integration of many topics or the use of very wide citation windows

Correlation between the number of top 0.1% cited papers and the corresponding Rn-index in several countries and research topics Country (type of papers) Topic top 0.1% Rn-index USA (collaborative) Stem cells 39 46.6 USA (domestic) Stem cells 31 58.8 EU (collaborative) Stem cells 29 32.1 USA (collaborative) Graphene 25 50.8 China (collaborative) Graphene 2...

work page 2018

[4] [5]

Conclusions The contribution to pushing the boundaries of knowledge is a critical metric for evaluating the research performance of countries and institutions, which in many cases is not revealed by common bibliometric indicators (Rodríguez-Navarro, 2024a; 2024b). The Rk-index was specifically designed to assess such contributions, and the Rn-index is a v...

work page Pith review arXiv 2013