@prefix this: . @prefix sub: . @prefix np: . @prefix rdf: . @prefix prov: . @prefix npx: . @prefix dc: . @prefix xsd: . sub:Head { this: a np:Nanopublication; np:hasAssertion sub:assertion; np:hasProvenance sub:provenance; np:hasPublicationInfo sub:pubinfo . } sub:assertion { a ; dc:description "Pre-registered. Load the CSV. Columns are per the record's field list: image-level fields (Dataset ID, Image Number, mean cell area, cell area CV, mean cell density, 4- to 9-sides cell counts, total cells, mean hexagonal regularity, number of pleomorphs, mean cell sides, cell sides CV, shape index) repeated on each per-cell row (cell area, number of sides, hexagonal regularity, shape index). Collapse to one row per image (Dataset ID + Image Number). Exclude images with fewer than 50 cells. Per image: f6 = 6-sides count / total cells; CV_area = the image's cell area CV as given; density = mean cell density. Primary statistic: Spearman rho between CV_area and f6 across images, with a 95% bootstrap CI (2,000 resamples). Repeat with images aggregated to one row per Dataset ID (mean of f6 and CV_area over that cornea's images) because images from the same cornea are not independent. Secondary, not deciding: Spearman rho between density and f6 (the clinical reading: cell loss goes with lost hexagonality); within-image Spearman rho between per-cell area and per-cell side count, median over images (Lewis's law, for comparison with ch07-09); sensitivity run restricted to images with density >= 2000 cells/mm^2 to check that the result is not driven by poor-quality or badly damaged tissue."; "Measured: -0.337 Spearman rho between per-image cell-area CV and per-image hexagonal fraction (n = 2,000 images, 354,998 cells (678 corneas per the record; not identifiable in the file), bootstrap 95% CI [-0.377, -0.297], 2,000 resamples). Interpretation: "; "Claim test of thompson1917-ch07-04 (eye-bank-hexagonality): refutes, confidence 0.55"; ; prov:wasAttributedTo ; "Planned datasets: auckland-nz-eye-bank-voronoi-endothelium-2017"; "claim-test/0.1.0"; "biological"; "Hexagonal fraction falls as cell-area CV rises: rho <= -0.30, at image and cornea level."; "A measurement confound is possible in principle: a poorer image could inflate area CV and lower the hexagon count together. The sensitivity run restricted to density >= 2,000 cells/mm^2 (1,850 images) gives rho -0.32 [-0.36, -0.28], so the result does not depend on the most damaged tissue, but image quality itself is not in the file.", "Cornea-level replication not possible: the file has no cornea or donor identifier (see deviations). Images from the same cornea are not independent, so the effective n is below 2,000; the CI is computed over images and is too narrow by an unknown factor. With 678 independent corneas the same rho would still clear the criterion, but that was not measured.", "Eye-bank corneas are pre-screened for transplant suitability, so density and CV are range-restricted (density p10-p90: 2,117-3,489 cells/mm^2; CV 15.5-20.2%). The weak density-hexagonality correlation (0.17) says nothing about diseased or very old endothelium.", "Side counts in a Voronoi tiling are a topological reading of 'packing'. The 120-degree force balance proved in ThompsonLaws.Cells.tripleJunction_120 is about angles and is untouched by this result; what is refuted is the claim that packing regularity is independent of cell size.", "Voronoi segmentation (Brookes 2017) imposes straight walls and three-way vertices; it can report sides, not junction angles, and cannot test the 'by threes' half of the claim. Mean sides of 5.998 shows the tiling is trivalent by construction."; "Hexagonal fraction is flat across the range of cell-area CV in non-dividing human corneal endothelium: |rho| < 0.15."; "Pre-registered. Refuted if rho(CV_area, f6) <= -0.30 at both the image and the cornea level with the upper 95% CI bound below -0.15 (more size variation goes with fewer hexagons even in the tissue where Thompson's condition is best met). Any other outcome, including the two levels disagreeing, is 'mixed'. If the CSV does not have the stated columns or image identifiers cannot be recovered, 'inconclusive'."; "0.55"^^xsd:double; ; "refutes"; "If junction geometry is independent of cell size in a non-dividing epithelium, then across 2,000 specular images of 678 organ-cultured human donor corneas the per-image hexagonal fraction is uncorrelated with the per-image coefficient of variation of cell area. The clinical reading (specular microscopy reports falling hexagonality together with rising area CV as cell loss) predicts a strong negative correlation instead."; "Spearman rank correlation with bootstrap CI at two levels of aggregation; no model fitting."; "Pre-registered. The 'irrespective of size' reading is supported if |rho(CV_area, f6)| < 0.15 at both the image level and the cornea level, with the bootstrap 95% CI inside (-0.30, 0.30)."; "true"^^xsd:boolean . . a ; dc:description "Homo sapiens, corneal endothelium, 354,998 cells in 2,000 images from 678 corneas of 351 donors, Voronoi-segmented after organ-culture storage"; dc:license "CC BY 4.0"; "Dataset: Measurements of ex vivo human corneal endothelium using Voronoi segmentation (Brookes, University of Auckland / New Zealand National Eye Bank, 2017)"; prov:atLocation ; prov:wasDerivedFrom ; "sha256:33a3ba27d6cd47b03d5882d2bdfa8edca9f78b1210eae86659b20d5331018f8d" . } sub:provenance { a prov:SoftwareAgent; "delphisci claim tester (claude-fable-5.1; cursor local session, corneal follow-up to the deck review)"; prov:actedOnBehalfOf ; . a ; dc:identifier "claude-fable-5.1" . sub:assertion prov:wasAttributedTo ; prov:wasDerivedFrom ; prov:wasGeneratedBy sub:test . sub:plan prov:hadPlan . sub:test a prov:Activity; prov:qualifiedAssociation sub:plan; prov:wasAssociatedWith . } sub:pubinfo { this: dc:created "2026-11-08T16:24:31.898Z"^^xsd:dateTime; dc:creator ; dc:license ; npx:hasNanopubType ; npx:introduces ; "Claim test of thompson1917-ch07-04 (eye-bank-hexagonality): refutes, confidence 0.55"; . sub:sig npx:hasAlgorithm "RSA"; npx:hasPublicKey "MIIBIjANBgkqhkiG9w0BAQEFAAOCAQ8AMIIBCgKCAQEAyFHtPB6p0mCjDozxlOY9CmO3nBm7Te3jFbXyLmM+G1tGe47Da0jYIbhkl9Oo1ZjxmCa3LCwlEqK8gj6hEIcL4ahW30JXh9HycfKVk1bugkzMEV15tJdTzAU3sGMFnbNsn43xY2LFLy45tApkRm5byEiM7oRRrooIdwoQKnP4Wzq/bfAoJ9qvW+MGkelyOTd1jbSt+KfmT7xKmDfWzAHddUlWwMX2QJOLxDaffOteTop6ZRgPiNmPJKA/S62Gac9XtSYT3oDkZ6/nr5cxxNLpOyxTf5FPISiCe11xYO/unYjY+mwnOfA8yhJIyx1l0O9I0vpOpn+E5TvyHJV+oSTrvwIDAQAB"; npx:hasSignature "GxdSMxAcavrpEvPc7420yLbp8Zy/9lAqUt9eUNwN3BgpopxMmLW8cKdfTIUvCS1b0uvqlJGzD+w6pgobRjXSpvvMDql1VI0YLfm7ZiYlzjKJ7+qwcpHANRnktAZX0i3tGBbnviJNlyxPyqP7yFaDkmu83anCoTjYsw3+o7UgjMMHEyMvg/7MYjLV2Z0jtWgg5wh+w+8VpH/k9ECKvTMwbiiyOPkNQ7uhgcqZSPro+Bh8MNWrK/dw35p43ZhHwSXexCiSpzj++y+7Nw5gAVPdWcFjxArJMxkaRaZpPvFgJElClmDJUlTE4JVImiehpUks/yKv9qj5riLG8OHeog5dSg=="; npx:hasSignatureTarget this:; npx:signedBy . }