Device-Independent Image Descriptors of the Landscape Montage Technique — public data and code bundle

Overview This bundle contains the analysis code and de-identified derived data supporting an ongoing study by Maeda, Okita, and Kato (2026) on device-independent image descriptors of the Landscape Montage Technique (LMT) and their association with Big Five personality and State-Trait Anxiety. It is intended to let readers inspect the analysis pipeline and independently verify the deposited descriptor matrices and benchmark summary tables reported in the accompanying manuscript, including the label-permutation global-null benchmark, the spatial-composition analysis with its robustness checks (threshold sweep, between-family test, distance-only subfamily, feature ablation, leave-one-out, leave-two-out, and club-stratified permutations), the mask-perturbation robustness check, and the Monte Carlo power evaluation of the proposed confirmatory design. New in version 3: the complete 168-pair spatial-composition correlation table and per-leave-out stability file; additional sensitivity analyses (leave-two-out and club-stratified permutation); the paint-skip extraction script and the Monte Carlo power-evaluation script with its results table; corrected README and citation metadata; English-valued summary tables; and a POSIX-compatible checksum manifest. Contents Analysis scripts (Python, 11 files): feature extraction from LMT drawings and their element masks (hue-area ratios from k-means representative colors, 1/f^n fluctuation exponents beta, area ratios, paint-skip ratios, and 24 spatial-composition descriptors), 336 Spearman correlations with multiplicity corrections, covariance-preserving permutation benchmarks, robustness and sensitivity analyses, a mask-perturbation check, and the Monte Carlo power evaluation of the confirmatory design. Derived data (CSV, 20 files): per-participant feature matrices with anonymized identifiers (P01 to P16), the full 336-pair and 168-pair correlation tables, permutation-benchmark and sensitivity summary tables, and aggregate-level descriptive statistics of the psychological variables (means, SDs, and ranges only; no per-participant scores). Figures (PNG, 5 files): results figures reproducible from the deposited data, except that the scatter figure additionally requires the withheld questionnaire scores and displays within-sample ranks only. The manuscript's pipeline schematic is an authors' illustration and is not reproduced here. Documentation: README, LICENSE and LICENSE-DATA, CITATION.cff, seeds.md (random seeds and determinism settings), and MANIFEST.sha256 (SHA-256 hash of every file; LF line endings, verifiable with sha256sum -c). Not included, and why Original drawings, hand-drawn element masks, and raw or totaled Big Five and STAI questionnaire responses are not deposited: they are sensitive projective and psychological material from a small cohort with residual re-identification risk, and only the minimum the scientific record requires is disclosed, consistent with the data-protection plan of the approving ethics protocol (no. 2024-4, Ethics Review Committee of the Graduate School of Information Sciences and Arts, Toyo University). They are available from the corresponding author under a data-sharing agreement. Licenses Analysis scripts: MIT License (see the LICENSE file). Derived data files and figures: Creative Commons Attribution 4.0 International (CC BY 4.0) (see the LICENSE-DATA file). Contact Corresponding author: Satoshi Maeda — [e-mail] — Department of Electrical, Electronic and Communications Engineering, Faculty of Science and Engineering, Toyo University.