{"id":"a4e637ed-45a5-41c9-8388-0000ee41586b","arxiv_id":"1908.06029","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":3.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"Pearson distance (1 minus the correlation) is not a metric, while its square-root form is; the same applies to the absolute-correlation versions.","lead":"Pearson distance, defined as one minus a correlation coefficient, is not a true metric because it can violate the triangle inequality. The paper shows that taking square roots restores the metric property, both for the ordinary and the sign-invariant versions.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Result III's proof that sqrt(1-rho^2) is a metric contains an invalid factorization: E(X_iY_iX_jY_j) does not equal E(X_iY_i)E(X_jY_j), since E(X_iY_i)=0 under the stated independence; the intended identity E(X_iX_j)E(Y_iY_j)=rho_ij^2 repairs it.","rationale":"The paper's main mathematical conclusions are correct: Pearson dissimilarity 1-rho and its absolute-value variant 1-|rho| violate the triangle inequality, while sqrt(1-rho) and sqrt(1-rho^2) are metrics. The negative results rest on explicit valid counterexamples, and the proof of Result I is a standard L2 triangle inequality argument after centering and scaling. The only real flaw I find is in Result III, where the displayed factorization is wrong: E(X_i Y_i X_j Y_j) is not E(X_i Y_i) E(X_j Y_j) because E(X_i Y_i)=0 under the stated independence. The correct factorization, E(X_i X_j) E(Y_i Y_j), yields rho_ij^2 and immediately repairs the proof. This is a proof gap, not a false theorem. The reader's weakest_assumption was the unstated finite-second-moment/nonzero-variance regularity; that is a worthwhile clarification but not a threat to the central claims. The reader did note the Result III typo in the rationale, so our concern partially overlaps. Since the error is minor and repairable, and the verdict CONDITIONAL already calls for such correction, no change to the verdict is warranted.","tokens_in":4359,"tokens_out":15845,"duration_ms":150481,"concrete_test":"Recompute the correlation of Z_i = X_i Y_i and Z_j = X_j Y_j in the exact setup of Result III: take independent bivariate Gaussian vectors X and Y with zero mean, unit variance, and corr(X_i, X_j) = rho_ij = 0.5. Compute E(Z_i Z_j) both symbolically and by Monte Carlo. The correct value is E(X_i X_j) E(Y_i Y_j) = 0.25, matching the repaired factorization; the printed factorization E(X_i Y_i) E(X_j Y_j) would give 0. This settles that the intended proof is valid once the equality is corrected.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The positive claim that P2earson dissimilarity sqrt(1-rho^2) is a metric rests on Result III. There, X and Y are independent random vectors with zero mean, unit variances, and common correlation matrix rho_ij. For i != j, the proof writes E(Z_i Z_j) = E(X_i Y_i X_j Y_j) = E(X_i Y_i) E(X_j Y_j) = rho_ij^2. This equality is false as written: because X and Y are independent vectors, E(X_i Y_i) = E(X_i) E(Y_i) = 0 and similarly E(X_j Y_j) = 0, so the printed product is 0, not rho_ij^2. The correct derivation is E(X_i Y_i X_j Y_j) = E(X_i X_j) E(Y_i Y_j), using independence of the X-vector and Y-vector, which gives rho_ij * rho_ij = rho_ij^2. Since this displayed step is the only justification for corr(Z_i, Z_j) = rho_ij^2, the proof as written has a genuine gap. The gap is easily repairable and the underlying claim is true, but the written proof requires correction. The counterexample-based negative claims (Result II and Result IV) are sound, as are the elementary proof of Result I after normalization.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper studies four correlation-based dissimilarity measures between random variables: Pearson dissimilarity 1-ρ_xy, |P|earson dissimilarity 1-|ρ_xy|, sqrt-Pearson dissimilarity √(1-ρ_xy), and P2earson dissimilarity √(1-ρ_xy^2). The main claims are that 1-ρ_xy and 1-|ρ_xy| are not metrics, while √(1-ρ_xy) and √(1-ρ_xy^2) are metrics. The negative results are supported by explicit counterexamples built from a latent variable construction (Results II and IV). The positive result for √(1-ρ_xy) is proved by an elementary L2 triangle-inequality argument (Result I). The positive result for √(1-ρ_xy^2) is proved by constructing Z_k = X_kY_k from two independent random vectors with identical correlation matrices and then applying Result I (Result III). The appendix sketches general results from Gower and Legendre (1986) on similarity matrices and from the metric-preserving function literature.","tokens_in":4668,"tokens_out":16368,"duration_ms":153080,"significance":"If the proof issue in Result III is corrected, the paper's main conclusions are correct and practically useful: it gives explicit, easily reproducible counterexamples showing that Pearson dissimilarity can violate the triangle inequality by large margins, and it identifies simple alternatives that are genuine metrics. The elementary proofs are accessible to a broad statistical audience, and the paper correctly connects the √(1-ρ) result to the older work of Gower and Legendre. The explicit counterexamples are falsifiable and parameterized over large ranges, which strengthens the message. The main weakness is the incorrect factorization in the proof of Result III, which is load-bearing for the claim that √(1-ρ^2) is a metric; this is repairable but must be fixed.","major_comments":[{"comment":"The displayed computation of the correlation of Z_i and Z_j contains an invalid factorization. The paper writes E(Z_iZ_j) = E(X_iY_iX_jY_j) = E(X_iY_i)E(X_jY_j) = ρ_ij^2. This is wrong under the stated assumptions: since the X-vector and Y-vector are independent, E(X_iY_i) = E(X_i)E(Y_i) = 0 and likewise E(X_jY_j) = 0, so the product on the right is 0, not ρ_ij^2. The correct derivation is E(X_iY_iX_jY_j) = E(X_iX_j)E(Y_iY_j) = ρ_ij ρ_ij = ρ_ij^2, using independence of the two vectors. Because this step is the only justification that corr(Z_i,Z_j)=ρ_ij^2, the proof of Result III as written is incomplete. The underlying claim is true and the gap is easily repaired, but the proof must be corrected.","section":"Section 3, Result III proof"},{"comment":"The paper defines a dissimilarity as satisfying strict positivity with 'equality iff i = j', but then lists 1-|ρ_ij| as a dissimilarity. For X and Y with Y = -X (after centering and scaling), one has |ρ_xy| = 1, so 1-|ρ_xy| = 0 even though the variables are not equal. This contradicts the stated axiom (ii). The authors should either relax the axiom (for example, to equality iff |correlation| = 1, treating variables as equivalent up to sign and affine transformation) or explicitly acknowledge that 1-|ρ| is not a dissimilarity in the paper's strict sense. This issue affects the framing of Result IV.","section":"Section 1, definition of dissimilarity"},{"comment":"The appendix states that no strictly convex function of √(1-ρ_ij) or of √(1-ρ_ij^2) that passes through the origin is a metric, and claims this follows from Corazza's theorem M1. That theorem says a strictly convex function with g(0)=0 is not metric preserving, meaning there exists some metric for which g∘d violates the triangle inequality. It does not, by itself, imply that a counterexample exists within the restricted class of metrics of the form √(1-ρ_ij) or √(1-ρ_ij^2). Supporting Result II* requires an argument that for every such strictly convex g the construction of Results II/IV (or a similar one) yields a violation within the correlation class. As written, the appendix overstates what the cited theorem establishes.","section":"Appendix, Result II*"}],"minor_comments":[{"comment":"The proof assumes zero mean and unit variance without stating that this is without loss of generality. Since correlations and the dissimilarities 1-ρ and √(1-ρ) are invariant to standardization, the proof should explicitly say that the variables are standardized first; otherwise the identity E(X-Y)^2 = 2(1-ρ_xy) does not hold for variables with arbitrary means and variances.","section":"Section 1, Result I proof"},{"comment":"Section numbering is inconsistent: the introduction refers to 'section II', 'section III', and 'section IV', while the actual section headers are numbered 1, 2, 3, and 4. Use one numbering convention consistently.","section":"Throughout"},{"comment":"The text says 'This certainly holds for |θ| ≤ π/4', but θ is defined on (0, π/2); the intended statement is '0 < θ ≤ π/4'.","section":"Section 2, Result II proof"},{"comment":"The first sentence says 'We use the construction from Theorem 2.2'; this should refer to 'Result II', not 'Theorem 2.2'.","section":"Section 3, Result IV proof"},{"comment":"The manuscript contains numerous typographical errors and OCR artifacts, including 'New So uth Wales', 'clu ster', 'fuctions' in the Corazza reference, missing spaces after commas in the reference list, 'Not that' for 'Note that', and '1N' for '1, ..., N'. A thorough proofreading is needed.","section":"Throughout"},{"comment":"The notation for correlations is inconsistent: both ρ_xy and ρx,y appear. Please choose one form and use it consistently.","section":"Notation"}],"recommendation":"major_revision","confidential_remarks":"This is a short methodological note that will be of interest to practitioners using correlation-based distances for clustering. The paper's central message — Pearson distance is not a metric — is correct and well illustrated by the counterexamples. The main issue is the erroneous factorization in the proof of Result III; it is an honest mistake that is easy to repair, but it must be corrected before publication. The authors should also tighten the definition of dissimilarity for 1-|ρ| and avoid overstating the appendix result. There is no concern about novelty or scope, but the paper would benefit from a careful proofreading pass."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick read of Solo's note. The core message is correct: 1 - rho is not a metric, and sqrt(1 - rho) and sqrt(1 - rho^2) are. The paper does a service by making that explicit for applied fields like gene expression and brain imaging that routinely treat Pearson distance as a metric. The counterexamples in Results II and IV are clean and show gross violations, not edge cases. The elementary proof that sqrt(1 - rho) is a metric is fine, a standard Minkowski/Cauchy-Schwarz argument after normalization. Now the soft spots. Most of the mathematical content is repackaged. As the appendix concedes, Gower and Legendre (1986) and Corazza (1999) already imply the non-metricity of 1 - rho and the metricity of the square-root versions. So the actual novelty is the explicit counterexample and the elementary exposition, not the theorem. That is worth something, but the paper should frame itself as a correction/tutorial note rather than a new result. The bigger issue is Result III. The proof that sqrt(1 - rho^2) is a metric contains an invalid factorization: it writes E(Z_i Z_j) = E(X_i Y_i) E(X_j Y_j), which is 0 under the stated independence. The intended identity is E(X_i Y_i X_j Y_j) = E(X_i X_j) E(Y_i Y_j) = rho_ij^2, using independence of the X and Y vectors. The gap is real but immediately repairable; the claim itself is true and is also covered by the appendix. This needs a correction. Minor: the paper never states the regularity condition that correlations require finite second moments and nonzero variances. The proofs assume zero mean and unit variance, so the statement should say that explicitly. There are also a few copyediting typos, like 'P2earson' in the list, that don't affect the math. Overall, this is a useful correction note for practitioners. I would accept it for peer review in an applied statistics or methods journal, with the expectation that the author either upgrades the novelty claim or repositions the note as an accessible proof of known facts. The explicit counterexample is a nice teaching tool, and the paper deserves a serious referee despite the flaws.","headline":"Pearson distance is indeed not a metric, and the paper's explicit counterexamples are useful, but the mathematical message is mostly repackaged known results and one proof contains a repairable gap.","tokens_in":694,"tokens_out":797,"would_cite":true,"duration_ms":28473,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["62H20","62H30"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that Pearson distance, defined as one minus the Pearson correlation coefficient, is not a metric because it can violate the triangle inequality, while its square root and the square root of one minus the squared…","keywords":["Pearson distance","correlation dissimilarity","metric","triangle inequality","clustering","metric-preserving functions","random variables"],"falsifier":"Compute the paper's construction at $\\theta=\\pi/6$: let $U,V,W$ be zero-mean with $\\mathrm{var}(V)=\\mathrm{var}(W)=1/4$, $\\mathrm{var}(U)=3/4$, $\\mathrm{corr}(V,W)=-3/4$, and $U$ independent of $V,W$; set $X=U+V$, $Y=U+W$, $Z=U$. Then $\\rho_{xy}=\\cos^4(\\pi/6)=9/16$, $\\rho_{xz}=\\rho_{yz}=\\cos(\\pi/6)=\\sqrt3/2$, so $1-\\rho_{xy}=7/16\\approx0.4375$ while $2(1-\\rho_{xz})=2-\\sqrt3\\approx0.2679$, violating the triangle inequality. Directly computing these correlations settles whether the paper's central claim is right.","tokens_in":4163,"feed_emoji":"📊","tokens_out":12001,"duration_ms":99126,"temperature":0.7,"pith_summary":"Pearson distance, commonly written $1-\\rho_{xy}$ and used in clustering gene-expression, brain-imaging, and cyber-security data, looks like a distance between random variables but is not one in the mathematical sense. The paper proves this by constructing three random variables for which the triangle inequality fails: $X$ and $Y$ are both close to $Z$ in Pearson distance, yet $X$ and $Y$ are farther apart than the sum of those two distances. It then shows the defect is repairable: $\\sqrt{1-\\rho_{xy}}$ and $\\sqrt{1-\\rho_{xy}^2}$ are metrics, while $1-|\\rho_{xy}|$ is not. Applications that assume Pearson distance obeys the triangle inequality are therefore relying on a false premise, and the square-root variants offer a drop-in fix.","feed_headline":"Pearson distance breaks the triangle inequality","feed_subtitle":"Clustering with one minus correlation can merge far-apart variables; the square root restores the metric.","key_machinery":"The engine of the positive results is the identity $\\mathbb{E}(X-Y)^2=2(1-\\rho_{xy})$ for zero-mean unit-variance variables, which identifies $\\sqrt{1-\\rho_{xy}}$ with the $L^2$ distance up to the constant $1/\\sqrt2$, so the $L^2$ triangle inequality transfers directly. For the sign-invariant metric, the paper uses a product construction: if $X$ and $Y$ are independent vectors with the same correlation matrix, then the products $Z_i=X_iY_i$ have correlations $\\rho_{ij}^2$, so $\\sqrt{1-\\rho_{ij}^2}$ is the $L^2$ distance of those products. The non-metric results rest on an explicit three-variable construction that reduces the triangle inequality to a trigonometric inequality, and the generalization section applies two metric-preserving-function rules: strictly convex functions through the origin do not preserve metrics, while strictly increasing concave functions through the origin do.","core_discovery":"The central claim is that the dissimilarity $1-\\rho_{xy}$ fails the metric axiom. The counterexample uses zero-mean variables $X=U+V$, $Y=U+W$, $Z=U$, where $V$ and $W$ have variance $\\sin^2\\theta$ and correlation $-\\cos^2\\theta$, $U$ has variance $\\cos^2\\theta$, and $X,Y$ end up with unit variance. This yields $\\rho_{xy}=\\cos^4\\theta$ and $\\rho_{xz}=\\rho_{yz}=\\cos\\theta$, so the triangle inequality for Pearson distance becomes $1-\\cos^4\\theta > 2(1-\\cos\\theta)$, which holds for a wide range of $\\theta$, including all $\\theta\\le\\pi/4$. The same construction shows $1-|\\rho_{xy}|$ is also not a metric. In the positive direction, $\\sqrt{1-\\rho_{xy}}$ is a metric because $\\mathbb{E}(X-Y)^2 = 2(1-\\rho_{xy})$ and the $L^2$ norm satisfies the triangle inequality, and $\\sqrt{1-\\rho_{xy}^2}$ is a metric by a product-copy argument that reduces it to the first case.","pith_inferences":["The paper's counterexample is not a razor-thin edge case: the violation holds for all $\\theta \\le \\pi/4$ and beyond, so for smoothly varying data one should expect triangle-inequality failures of Pearson distance to be common rather than rare in practice.","A testable extension is to audit real datasets, such as gene-expression or multi-voxel brain-imaging inputs, for triangle-inequality violations under Pearson distance; if violations are frequent, the square-root replacement would change cluster assignments in a way current practice misses.","The product-copy argument indicates that the metric property of $\\sqrt{1-\\rho^2}$ should hold empirically for any correlation matrix estimated from data with finite variances, independent of the underlying distribution.","The family of concave transformations opens a modelling choice: different concave powers compress large dissimilarities differently, which changes cluster shape, so selecting among them is a substantive decision rather than a technicality."],"forward_implications":["Clustering pipelines that use Pearson distance as a dissimilarity measure can produce incoherent clusters, because the triangle inequality is not guaranteed: two variables can both be near a third while being far from each other.","Replacing Pearson distance with $\\sqrt{1-\\rho_{xy}}$ restores the metric property while preserving the ordering of pairwise dissimilarities, so users only need to apply a square root to their existing matrix.","For analyses that treat anti-correlation as similarity, $1-|\\rho_{xy}|$ has the same defect, and $\\sqrt{1-\\rho_{xy}^2}$ is the metric replacement.","The metric-preservation results imply that further transformations, such as $(1-\\rho_{xy})^{1/4}$, are also metrics, giving a family of correlation-based distances with different geometrical behaviour.","Because the proofs require only finite second moments, the metric and non-metric conclusions apply to any collection of real-valued random variables with well-defined correlations, not just to specific distributions."],"supporting_citations":[{"why":"Supplies the informal result that $\\sqrt{1-s_{ij}}$ is a metric when the similarity matrix is positive semidefinite, the historical basis for the square-root repair.","marker":"Gower (1966)"},{"why":"Proves the formal metric result for $\\sqrt{1-s_{ij}}$ under positive semidefiniteness; the appendix uses it to derive Result I and frame the general approach.","marker":"Gower and Legendre (1986)"},{"why":"Provides the metric-preserving-function results M1 and M2 that the paper uses to show $1-\\rho_{xy}$ and $1-|\\rho_{xy}|$ are not metrics and that concave transformations preserve metricity.","marker":"Corazza (1999)"},{"why":"Original source for the claim that $\\sqrt{1-\\rho_{ij}^2}$ is a metric; the paper credits this result and gives an elementary proof.","marker":"Innocenti and Materassi (2008)"}],"fun_headline_variants":["Pearson distance violates the triangle inequality","Square root rescues Pearson distance","Pearson distance: one square root short of a metric","Clustering with Pearson distance? Check the triangle inequality","Pearson distance isn't a true metric"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that every random variable involved has a well-defined Pearson correlation, which requires finite second moments and nonzero variance; the paper normalizes to zero mean and unit variance without stating this regularity condition explicitly.","fun_headline_variants_meta":{"raw":{"variants":["Pearson distance violates the triangle inequality","Square root rescues Pearson distance","Pearson distance: one square root short of a metric","Clustering with Pearson distance? Check the triangle inequality","Pearson distance isn't a true metric"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001596,"raw_usage":{"total_tokens":6357,"prompt_tokens":940,"completion_tokens":5417,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":556,"completion_tokens_details":{"reasoning_tokens":5349}},"tokens_in":556,"tokens_out":5417,"duration_ms":35981,"temperature":1.0,"reasoning_tokens":5349,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T13:14:35.372211+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the paper's construction at $\\theta=\\pi/6$: let $U,V,W$ be zero-mean with $\\mathrm{var}(V)=\\mathrm{var}(W)=1/4$, $\\mathrm{var}(U)=3/4$, $\\mathrm{corr}(V,W)=-3/4$, and $U$ independent of $V,W$; set $X=U+V$, $Y=U+W$, $Z=U$. Then $\\rho_{xy}=\\cos^4(\\pi/6)=9/16$, $\\rho_{xz}=\\rho_{yz}=\\cos(\\pi/6)=\\sqrt3/2$, so $1-\\rho_{xy}=7/16\\approx0.4375$ while $2(1-\\rho_{xz})=2-\\sqrt3\\approx0.2679$, violating the triangle inequality. Directly computing these correlations settles whether the paper's central claim is right.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Proves the formal metric result for $\\sqrt{1-s_{ij}}$ under positive semidefiniteness; the appendix uses it to derive Result I and frame the general approach."},{"cited_title":"(1999), ‘Introduction to metric preserving fuctions ’, Amer","cited_arxiv_id":null,"evidence_quote":"Provides the metric-preserving-function results M1 and M2 that the paper uses to show $1-\\rho_{xy}$ and $1-|\\rho_{xy}|$ are not metrics and that concave transformations preserve metricity."}],"review_version":1}