The exact region between Chatterjee's and Blest's rank correlations
International Journal of Approximate Reasoning · DOI 10.1016/j.ijar.2026.109744
The exact xi-Blest region, both explicit coefficient branches, endpoint limits, unique curved-boundary copulas and parameters, convexity, compactness, and the sharp gap 44/105 are checked. The relaxed measurable-kernel optimization problem, normalization-map continuity, and derivative identity including b=1 are also checked. Mapped arXiv v1 auxiliary results include Lemmas 4.1-4.2. Two printed arXiv v1 errors are formally refuted: an intermediate polynomial and the direction of the shuffling lemma's concordance order. The corrected polynomial is verified. The author revision's signed-family concordance ordering, coefficient reflection, negative-parameter SD, uniform copula limits, positive-parameter MTP2 density existence, and the source density candidate's switch points and inverse-normalization coefficient are checked. The revised lemma's displayed derivative-form density is identified with the actual copula measure for every positive parameter. The corrected four-part shuffling lemma is verified, including its xi invariance, decreasing concordance order and quantitative uniform CDF continuity. All original named results in the stated scope are matched to the final IJAR PDF in SOURCE_COMPARISON.md; externally cited lemmas, numerical tables, and figures are outside the formal scope.
Verification map
Status: complete for scope. The exact xi-Blest region, both explicit coefficient branches, endpoint limits, unique curved-boundary copulas and parameters, convexity, compactness, and the sharp gap 44/105 are checked. The relaxed measurable-kernel optimization problem, normalization-map continuity, and derivative identity including b=1 are also checked. Mapped arXiv v1 auxiliary results include Lemmas 4.1-4.2. Two printed arXiv v1 errors are formally refuted: an intermediate polynomial and the direction of the shuffling lemma's concordance order. The corrected polynomial is verified. The author revision's signed-family concordance ordering, coefficient reflection, negative-parameter SD, uniform copula limits, positive-parameter MTP2 density existence, and the source density candidate's switch points and inverse-normalization coefficient are checked. The revised lemma's displayed derivative-form density is identified with the actual copula measure for every positive parameter. The corrected four-part shuffling lemma is verified, including its xi invariance, decreasing concordance order and quantitative uniform CDF continuity. All original named results in the stated scope are matched to the final IJAR PDF in SOURCE_COMPARISON.md; externally cited lemmas, numerical tables, and figures are outside the formal scope.
Source and conventions
Source: arXiv:2603.09768v1, 10 March 2026. Journal reference: DOI 10.1016/j.ijar.2026.109744. Selected published statements are compared in SOURCE_COMPARISON.md.
Blest weights the first coordinate by 1-u, exactly as in equation (3). The product uniform measure is proved to give the displayed iterated integral. All integrals exist for arbitrary copulas, including singular ones. Xi conditions coordinate 1 on coordinate 0, with the derivative bridge checked in equation (2). Pi/M/W benchmarks and the full range [-1,1] are checked. Reflection of the second coordinate preserves xi and negates nu. The extremal construction and optimization results are mapped separately below. No sign-reversal claim is made for reflection of the first coordinate.
The vertical boundary is obtained using the proved identity nu=rho for radially symmetric copulas and centered deterministic witnesses with rho=1-2 alpha^3. The clamped family and its sharp optimization certificate are now proved; both coefficient branches are now derived from exact integrals of the constructed family.
Convexity is proved independently of the curved boundary formulas. A first copula mixture preserves the desired affine coefficient and gives xi no larger than the target convex combination. A second mixture with a proved xi=1 witness at the same coefficient reaches the target xi by continuity. Both steps construct actual copulas, including singular laws and endpoint weights. This also proves attainment of every xi between an existing point and 1 at fixed coefficient; closedness and attainment of slice extrema are proved separately below.
Result map
Main.lean imports the result modules. Axioms.lean prints and enforces the standard transitive axiom allowlist for every declaration below.
The coefficient computation uses a uniform sampling representation, exact clamp-noise moments, and radical integrals. This is an independent derivation of Theorem 2.3; the separate section and substitution formulas are also verified below. The relaxed optimization theorem covers all measurable representatives satisfying the box and marginal constraints almost everywhere, without a monotonicity assumption. Uniqueness is equality almost everywhere, as appropriate for L2.
| Source result | Lean declaration | Status | Hypotheses and scope |
|---|---|---|---|
| Equation (3): existence of the weighted integral | Papers.Rockel2026XiBlest.blest_integrable |
verified | All bivariate copulas; no density hypothesis. |
| Equation (3): iterated-integral correspondence | Papers.Rockel2026XiBlest.blest_integral_formula |
verified | First-coordinate weight 1-u and source integration order. |
| Equation (2): xi derivative convention | Papers.Rockel2026XiBlest.xi_derivative_formula |
verified | All bivariate copulas; almost-everywhere derivative correspondence. |
| After equation (3): concordance monotonicity | Papers.Rockel2026XiBlest.blest_mono |
verified | Pointwise CDF ordering implies ordering of Blest coefficients. |
| Equation (3): affine-mixture identity used in the region argument | Papers.Rockel2026XiBlest.blest_mix |
verified | Any two copulas and every closed-interval mixture weight. |
| After equation (3): independence normalization | Papers.Rockel2026XiBlest.blest_independence |
verified | Nu(Pi)=0 with the source normalization. |
| Theorem 1.1: xi=0 endpoint | Papers.Rockel2026XiBlest.xi_zero_slice |
verified | Xi=0 forces the unique copula Pi and nu=0; both curved boundaries and the xi=1 boundary are checked below. |
| Supporting identity for Theorem 1.1 vertical boundary | Papers.Rockel2026XiBlest.blest_eq_rho_of_radiallySymmetric |
verified | Blest nu equals Spearman rho for every radially symmetric copula. Both uniform-coordinate reflection and weighted-integral identities are proved. |
| Theorem 1.1: entire vertical boundary | Papers.Rockel2026XiBlest.xi_one_slice |
verified | Exactly nu in [-1,1] is attainable at xi=1. Centered deterministic radially symmetric copulas supply the witnesses; no unproved curved-boundary formula is used. |
| Theorem 1.1: attainment up to xi=1 at fixed Blest nu | Papers.Rockel2026XiBlest.fixed_coefficient_upward |
verified | Every attained (x,y) extends to all (z,y) with x<=z<=1. Uses an actual xi=1 witness at the same coefficient and a continuous copula-mixture path. |
| Theorem 1.1: convexity of the entire attainable region | Papers.Rockel2026XiBlest.attainable_region_convex |
verified | Every convex combination of attainable pairs is attained. Independent two-mixture proof; no assumed curved-boundary formula or full-region characterization. |
| Equation (3): reversed integration order | Papers.Rockel2026XiBlest.blest_integral_formula_swapped |
verified | Fubini and integrability for the weighted CDF of every copula. |
| After equation (3): M/W normalization and range | Papers.Rockel2026XiBlest.blest_comonotonic; Papers.Rockel2026XiBlest.blest_countermonotonic; Papers.Rockel2026XiBlest.blest_mem_Icc |
verified | Nu(M)=1, nu(W)=-1, and -1<=nu(C)<=1 for every copula. |
| Equation (13) and the following xi calculation | Papers.Rockel2026XiBlest.blest_reflect_second; Papers.Rockel2026XiBlest.xi_blest_reflection |
verified | Reflecting coordinate 1 preserves xi and negates Blest's nu, including singular copulas. |
| Theorem 1.1: continuity of Blest under CDF convergence | Papers.Rockel2026XiBlest.blest_tendsto_of_cdf |
verified | Pointwise CDF convergence of arbitrary copulas implies convergence of the exact weighted functional. |
| Theorem 1.1: closedness and compactness | Papers.Rockel2026XiBlest.attainable_region_closed; Papers.Rockel2026XiBlest.attainable_region_compact |
verified | Entire attained region. Independent proof using weak compactness, lower semicontinuity of xi, and fixed-Blest upward interpolation. |
| Theorem 1.1: attained slice extrema | Papers.Rockel2026XiBlest.blest_extrema_attained; Papers.Rockel2026XiBlest.minimal_xi_attained |
verified | Both Blest extrema exist for each xi in [0,1]; the least xi exists for every Blest value in [-1,1]. The extremal family, formulas, and uniqueness are identified by the separate results below. |
| The final IJAR numbering and every original named result in the stated scope are compared in SOURCE_COMPARISON.md. The two refuted intermediate claims belong only to arXiv v1 and are corrected in the published PDF. | |||
| Equation (16): conditional integral identity | Papers.Rockel2026XiBlest.blest_conditional_formula |
verified | Every copula, including singular laws; the squared first-coordinate weight is proved by Fubini. |
| Lemma 2.1: unique normalization | Papers.Rockel2026XiBlest.extremal_normalization |
verified | Every b>0 and v in [0,1]; unique q in [-1/b,1]. |
| Lemma 2.1: ordered, measurable conditional sections | Papers.Rockel2026XiBlest.extremalQ_antitone; Papers.Rockel2026XiBlest.extremal_kernel_measurable |
verified | Antitone q and joint measurability of the actual normalized kernel. |
| Lemma 2.2: constructed copula and derivative bridge | Papers.Rockel2026XiBlest.extremal_cdf; Papers.Rockel2026XiBlest.extremal_conditionalCDF; Papers.Rockel2026XiBlest.extremal_isSI |
verified | The clamped kernel constructs an actual copula; CDF formula, almost-everywhere conditional identity, and stochastic monotonicity. |
| Theorem 1.1: sharp quantitative optimization certificate | Papers.Rockel2026XiBlest.clamped_blest_distance_bound; Papers.Rockel2026XiBlest.extremal_support; Papers.Rockel2026XiBlest.extremal_support_eq_iff |
verified | All competing copulas. The support deficit bounds squared conditional-CDF distance and equality forces the constructed copula. |
| Theorem 1.1: unique maximum at every constructed positive-slope point | Papers.Rockel2026XiBlest.extremal_maximal_blest; Papers.Rockel2026XiBlest.extremal_maximal_blest_eq_iff |
verified | Blest maximality for xi no greater than the constructed point; equality characterization at the same xi. Exhaustion of all intermediate xi values is proved below. |
| Theorem 1.1: independence endpoint of the constructed family | Papers.Rockel2026XiBlest.extremal_zero |
verified | The b=0 extension is exactly the independence copula. |
| Theorem 1.1: monotonicity of the extremal coefficients | Papers.Rockel2026XiBlest.extremal_coefficients_monotone |
verified | Both xi and Blest are nondecreasing in the nonnegative slope; strict parameter monotonicity is verified below. |
| Theorem 1.1: continuous xi parameter | Papers.Rockel2026XiBlest.extremal_xi_continuous |
verified | The entire nonnegative parameter interval, including zero; proved with the quantitative bound abs(xi(b)-xi(d))<=24 abs(b-d). |
| Theorem 1.1: comonotonic endpoint | Papers.Rockel2026XiBlest.extremalSequence_cdf; Papers.Rockel2026XiBlest.extremalSequence_xi; Papers.Rockel2026XiBlest.extremalSequence_blest |
verified | The cofinal sequence b=(n+1)^2 converges in CDF, xi, and Blest to the comonotonic endpoint. The uniform CDF error is at most 1/(n+1). |
| Theorem 1.1: all intermediate xi parameters are attained | Papers.Rockel2026XiBlest.extremal_parameter_exists |
verified | Every x in (0,1) is the xi of an actual positive-slope member; proved by continuity and the checked endpoint limit. |
| Theorem 1.1: exact attainable slices | Papers.Rockel2026XiBlest.extremal_slice_iff; Papers.Rockel2026XiBlest.intermediate_slice_characterization |
verified | For every intermediate xi, the full Blest slice is exactly the interval between the reflected constructed boundaries. Actual copula witnesses fill the interval; the upper boundary copula is unique. Both explicit coefficient branches are verified below. |
| Theorem 2.3: polynomial coefficient branch | Papers.Rockel2026XiBlest.extremal_coefficients_polynomial |
verified | Actual extremal copulas for every 0<=b<=1: xi=8b^2(7-3b)/105 and nu=4b(28-9b)/105, including b=0 and b=1. Derived through the proved uniform sampling law and exact integrals. |
| Theorem 2.3: joining point of the two branches | Papers.Rockel2026XiBlest.extremal_one_coefficients |
verified | At b=1, xi=32/105 and nu=76/105. |
| Theorem 1.1: maximal signed gap and uniqueness | Papers.Rockel2026XiBlest.maximal_signed_gap; Papers.Rockel2026XiBlest.maximal_signed_gap_eq_iff |
verified | For every copula, nu-xi<=44/105; equality holds exactly at the constructed b=1 copula. |
| Theorem 1.1: maximal absolute gap and all equality cases | Papers.Rockel2026XiBlest.maximal_absolute_gap; Papers.Rockel2026XiBlest.maximal_absolute_gap_eq_iff |
verified | For every copula, abs(nu)-xi<=44/105; equality holds exactly at the b=1 copula or its response reflection. |
| Lemma 4.5 and Theorem 1.1: strict coefficient monotonicity | Papers.Rockel2026XiBlest.extremal_coefficients_strictMono |
verified | Both xi and Blest strictly increase with the nonnegative slope. Proved from slope identifiability and the sharp optimization certificate, independently of the b>1 formulas. |
| Theorem 1.1: finite parameters and unique xi parameter | Papers.Rockel2026XiBlest.extremal_finite_xi; Papers.Rockel2026XiBlest.extremal_positive_xi; Papers.Rockel2026XiBlest.extremal_parameter_unique |
verified | Every positive finite slope has xi strictly between 0 and 1; every such xi has exactly one positive parameter. |
| Theorem 2.3: hyperbolic coefficient branch | Papers.Rockel2026XiBlest.extremal_coefficients_hyperbolic |
verified | For every b>1, both displayed arcosh formulas equal the coefficients of the constructed copula. Exact radical integrals; no assumed coefficient formula. |
| Theorem 2.3 and equations (4)-(5): complete formulas | Papers.Rockel2026XiBlest.extremal_coefficients |
verified | All b>=0, including the joining point and independence endpoint. |
| Theorem 1.1 and equation (6): explicit exact region | Papers.Rockel2026XiBlest.exact_region |
verified | Necessary and sufficient conditions, with actual copula witnesses for every point and the full xi=1 vertical segment. |
| Theorem 1.1: unique upper and lower curved boundaries | Papers.Rockel2026XiBlest.upper_boundary_unique; Papers.Rockel2026XiBlest.lower_boundary_unique |
verified | Every finite positive slope; the lower optimizer is precisely the response reflection. |
| Theorem 1.1: endpoint conventions and limits | Papers.Rockel2026XiBlest.formula_zero; Papers.Rockel2026XiBlest.formula_zero_limits; Papers.Rockel2026XiBlest.formula_infinity_limits |
verified | Both coefficients equal and tend to zero at b=0, and tend to one as the real parameter tends to infinity. |
| Lemma 4.5 and equation (7): derivative identity | Papers.Rockel2026XiBlest.formula_hasDerivAt; Papers.Rockel2026XiBlest.formula_derivative_identity |
verified | Both functions are differentiable for every b>0, including b=1, and N′(b)=Xi′(b)/b. Differentiation under the exact tail integrals is justified. |
| Lemma 2.1: normalization-map properties | Papers.Rockel2026XiBlest.normalizationMean_properties; Papers.Rockel2026XiBlest.extremalQ_continuous; Papers.Rockel2026XiBlest.extremal_kernel_continuous |
verified | Continuous, strictly decreasing normalization map on [-1/b,1], correct endpoints and range, continuous inverse parameter, and jointly continuous kernel. |
| Lemma 4.4: mixture-path continuity | Papers.Rockel2026XiBlest.xi_mixture_continuous |
verified | All copulas, including singular laws and endpoint mixture weights. |
| Theorem 3.4: relaxed optimization and uniqueness | Papers.Rockel2026XiBlest.relaxed_distance_bound; Papers.Rockel2026XiBlest.relaxed_solution |
verified | All admissible measurable kernels, including those not defining copulas. Every c in (0,1) has a unique positive parameter and a unique optimizer modulo almost-everywhere equality. |
| Lemma 4.1: equations (20)-(22) | Papers.Rockel2026XiBlest.section_formulas |
verified | All admissible section parameters, including clamp-switching endpoints; exact ordinary, squared, and weighted moments. |
| Lemma 4.1: differentiability and change of variables | Papers.Rockel2026XiBlest.normalizationMean_hasDerivAt; Papers.Rockel2026XiBlest.normalization_substitution |
verified | Differentiation under the integral is justified by a uniform Lipschitz bound and null switching sets. The substitution holds for every continuous section functional. |
| Lemma 4.1: equations (23)-(24) | Papers.Rockel2026XiBlest.one_dimensional_coefficients |
verified | The one-dimensional integrals equal the actual copula coefficients, using the derivative of the actual normalization map. |
| Lemma 4.2: all four regimes | Papers.Rockel2026XiBlest.substitution_upper; Papers.Rockel2026XiBlest.substitution_unclamped; Papers.Rockel2026XiBlest.substitution_double; Papers.Rockel2026XiBlest.substitution_lower |
verified | Upper-clamped, unclamped, double-clamped, and lower-clamped substitutions, including shared boundary cases. |
| Revised Lemma 4.3(i): corrected shuffle construction and endpoints | Papers.Rockel2026XiBlest.shufflePoint_involution; Papers.Rockel2026XiBlest.shuffleReal_ae_printed; Papers.Rockel2026XiBlest.shufflePoint_measurePreserving; Papers.Rockel2026XiBlest.shuffledCopula_zero; Papers.Rockel2026XiBlest.shuffledCopula_one; Papers.Rockel2026XiBlest.shuffleReal_ae_printed_copula |
verified | For any copula and p in [0,1], the endpoint-safe upper-interval flip preserves the uniform first marginal and defines an actual copula. Its laws at p=0 and p=1 are C and the first-coordinate reflection. It agrees almost everywhere with the revision's printed endpoint convention under every copula. |
| Revised Lemma 4.3(ii): xi invariance of the shuffle | Papers.Rockel2026XiBlest.copula_measureReal_conditional_set; Papers.Rockel2026XiBlest.shuffle_setIntegral; Papers.Rockel2026XiBlest.shuffledCopula_conditionalCDF_ae; Papers.Rockel2026XiBlest.shuffledCopula_xi |
verified | For every copula and every p in [0,1], xi of the actual shuffled copula equals xi of the original. An arbitrary-set disintegration identity and measure-preserving involution identify the transformed conditional CDF almost everywhere; singular copulas are included. |
| Revised Lemma 4.3(iii): corrected concordance order | Papers.Rockel2026XiBlest.shuffledCopula_cdf_below_cut; Papers.Rockel2026XiBlest.shuffledCopula_cdf_above_cut; Papers.Rockel2026XiBlest.si_cdf_cross; Papers.Rockel2026XiBlest.shuffledCopula_concordance |
verified | For every SI copula and p>=p', the actual shuffled copulas satisfy C_p<=_co C_p'. Exact CDF formulas on both sides of the split and a four-point SI inequality prove the reversed direction used in the local author revision. |
| Revised Lemma 4.3(iv): uniform path continuity | Papers.Rockel2026XiBlest.shuffledCopula_cdf_bound; Papers.Rockel2026XiBlest.shuffledCopula_uniformCDF_continuous |
verified | For every copula and p,p' in [0,1], the absolute CDF difference is at most 2 abs(p-p') at every point of the closed square. This gives uniform convergence as p tends to p' with a parameter-independent spatial bound; SI is unnecessary. |
| arXiv v1 Lemma 4.3(iii): refutation of printed shuffling order | Papers.Rockel2026XiBlest.not_comonotonic_concordanceLE_countermonotonic |
verified | The arXiv v1 direction would give M <=_co W at p=1, p'=0 with C=M. Lean disproves this endpoint case. The local author revision reverses the direction, and the full corrected shuffle path is verified in the three rows above. |
| arXiv v1 proof of Theorem 2.3: corrected lower-clamped square polynomial | Papers.Rockel2026XiBlest.lower_square_polynomial; Papers.Rockel2026XiBlest.printed_lower_square_polynomial_false |
verified | The correct coefficient of r^5 is -8/15. The printed -1/5 is disproved at r=1/2. The final theorem formulas are unaffected and separately verified. |
The verified subset consists only of the explicitly mapped statements and proof steps. Pending rows are not implied by a successful build. Numerical experiments and plots are not counted as formal proofs.
Lemma 3.1 is quoted external KKT theory; Lemmas 3.2 and 4.3 are quoted from the xi-rho paper. The arXiv v1 statement of Lemma 4.3(iii) has the concordance direction reversed, as the formal endpoint counterexample shows; all four corrected shuffling claims are now proved. The independent exact-region proof does not rely on the KKT or shuffling lemmas. Numerical Table 1 and plots are excluded from the formal theorem scope.
Corrected intermediate calculation
In arXiv v1, the lower-clamped expression G_iv in the proof of Theorem 2.3 subtracts r^5/5. Expanding the preceding definition F(r;r^2) instead gives 8r^5/15. At b=1 and r=1/2, the correct squared section integral is 19/240, whereas the printed expression gives 43/480. The two Lean declarations above prove the corrected identity and disprove the printed polynomial. The final Xi and N formulas and all region statements are proved from the actual copulas. The literal erroneous arXiv v1 intermediate line is not claimed verified. The final IJAR PDF instead prints the corrected -8r^5/15 term.
Additional revision obligations
The local author revision adds family properties beyond arXiv v1; see SOURCE_COMPARISON.md.
| Source result | Lean declaration | Status | Hypotheses and scope |
|---|---|---|---|
| Revised Lemma cb_copula: absolute continuity and concrete density witness | Papers.Rockel2026XiBlest.extremal_standardized_density; Papers.Rockel2026XiBlest.extremal_hasMTP2Density |
verified | The actual copula measure equals Lebesgue measure weighted by the standardized-band density. This gives absolute continuity and a measurable nonnegative density for every b>=0. The source's derivative form is verified below for b>0. |
| Revised Lemma cb_copula: inverse derivative and density coefficient | Papers.Rockel2026XiBlest.quadratic_switch_strict; Papers.Rockel2026XiBlest.extremalQ_interior; Papers.Rockel2026XiBlest.normalizationMean_deriv_neg_at_extremalQ; Papers.Rockel2026XiBlest.extremalQExtension_coe; Papers.Rockel2026XiBlest.extremalQExtension_hasDerivAt; Papers.Rockel2026XiBlest.extremalQExtension_density_coefficient |
verified | For every b>0 and interior v, the actual normalization inverse q has a nonzero derivative. The coefficient -b q'(v) equals the reciprocal length of the unclamped band. |
| Revised Lemma cb_copula: switching points and support | Papers.Rockel2026XiBlest.extremalLowerSwitch_source; Papers.Rockel2026XiBlest.extremalUpperSwitch_source; Papers.Rockel2026XiBlest.extremal_switch_support |
verified | Both source switch-point expressions are identified with the reflected endpoints of the active clamped-square band, including the open support inequalities. |
| Revised Lemma cb_copula: derivative-formula candidate | Papers.Rockel2026XiBlest.extremalDerivativeDensity_eq_width_of_interior; Papers.Rockel2026XiBlest.extremalDerivativeDensity_ae_eq_width; Papers.Rockel2026XiBlest.extremalWidthDensity_nonneg; Papers.Rockel2026XiBlest.extremalWidthDensity_measurable |
verified | The printed derivative expression equals a measurable, nonnegative reciprocal-width function almost everywhere; the excluded response endpoints have zero Lebesgue measure. |
| Revised Lemma cb_copula: standardized-band quantile and support bridge | Papers.Rockel2026XiBlest.extremal_raw_quantile_parameter; Papers.Rockel2026XiBlest.extremal_raw_band_condition; Papers.Rockel2026XiBlest.quadratic_active_interval_integral; Papers.Rockel2026XiBlest.quadratic_closed_active_reflected_integral; Papers.Rockel2026XiBlest.quadratic_active_iff_switch; Papers.Rockel2026XiBlest.extremal_raw_columnDensity; Papers.Rockel2026XiBlest.extremal_raw_support_ae |
verified | The raw-band quantile has intercept -b q(v); its column density is (b+1) times the active width. Its closed-strip support equals the source's open support almost everywhere, including the null switching boundaries. |
| Revised Lemma cb_copula: displayed a.e. density as the actual copula measure | Papers.Rockel2026XiBlest.extremal_standardized_eq_width_of_support; Papers.Rockel2026XiBlest.extremalWidthDensity_ae_standardized; Papers.Rockel2026XiBlest.extremal_toMeasure_widthDensity; Papers.Rockel2026XiBlest.extremal_toMeasure_derivativeDensity |
verified | For every b>0, the copula measure equals Lebesgue measure weighted by the revised manuscript's derivative-form density. The reciprocal-width and standardized-band forms agree almost everywhere. |
| Revised Proposition basic_properties_cb(i): MTP2 | Papers.Rockel2026XiBlest.extremal_hasMTP2Density; Papers.Rockel2026XiBlest.signed_extremal_hasMTP2Density |
verified | Every finite positive member has a measurable nonnegative Lebesgue density satisfying MTP2 on the entire square; also covers the independence extension at zero. A fixed-width increasing band is standardized through its actual second marginal, and its copula is proved equal to the clamped extremal family. |
| Revised Proposition basic_properties_cb(i): negative-parameter SD | Papers.Rockel2026XiBlest.signed_extremal_isSD |
verified | Every negative parameter gives an SD copula, using the source response reflection and the positive SI theorem. |
| Revised Proposition basic_properties_cb(ii): concordance order | Papers.Rockel2026XiBlest.extremal_cdf_monotone; Papers.Rockel2026XiBlest.signed_extremal_concordance |
verified | Actual CDF ordering for all signed parameters, including the independence extension at zero. Equal section means and the single crossing of affine quadratic profiles prove the positive ordering; reflection proves the negative branch. |
| Revised Proposition basic_properties_cb(iii): coefficient reflection | Papers.Rockel2026XiBlest.signed_extremal_reflection_coefficients |
verified | For every b>0, the actual signed family has equal xi at b and -b and opposite Blest nu. |
| Revised Proposition basic_properties_cb(iv): uniform endpoint limits | Papers.Rockel2026XiBlest.signed_extremal_independence_error; Papers.Rockel2026XiBlest.signed_extremal_uniform_zero; Papers.Rockel2026XiBlest.signed_extremal_uniform_top; Papers.Rockel2026XiBlest.signed_extremal_uniform_bot |
verified | Uniform convergence of the actual copula CDFs on the entire closed square to Pi, M and W as the real parameter tends to zero, positive infinity and negative infinity. The zero limit holds from both sides, with error at most 2 abs(b). |
Inspect the formalization
Reproduce this snapshot
Run the full project build to check every source file. The pinned toolchain and dependencies live at the repository root.
git clone https://github.com/Corrram/lean-verifications.git
cd lean-verifications
git checkout --detach 0ac8668a9c4c9007ec374694abc14296edbeae75
lake exe cache get
lake buildCite this supplement
Use the permanent folder at commit 0ac8668 to identify the exact software snapshot. Cite the original article separately. This handbook follows the latest deployed commit.
Article bibliography (BibTeX)
@article{Rockel2026XiBlest,
author = {Rockel, Marcus},
title = {The exact region between Chatterjee's and Blest's rank correlations},
journal = {International Journal of Approximate Reasoning},
year = {2026},
volume = {197},
pages = {109744},
doi = {10.1016/j.ijar.2026.109744},
url = {https://doi.org/10.1016/j.ijar.2026.109744}
}
@misc{Rockel2026XiBlestArxiv,
author = {Rockel, Marcus},
title = {The exact region between Chatterjee's and Blest's rank correlations},
year = {2026},
eprint = {2603.09768},
archivePrefix = {arXiv},
primaryClass = {math.ST},
note = {Version 1, 10 March 2026},
url = {https://arxiv.org/abs/2603.09768v1}
}
% Cite the Lean supplement separately using its folder and full commit permalink.