ARGIRA XI-A — GLCM Residual Structure Analysis of ARGIRA X

ARGIRA XI-A — GLCM Residual Structure Analysis of ARGIRA X ARGIRA XI-A is a reproducible dataset and analytical framework that investigates whether second-order texture descriptors derived from the Gray-Level Co-occurrence Matrix (GLCM) contain predictive information about the residual structure identified in ARGIRA X. The study operationalizes an explicit projection from image feature space into the residual vector space εₓ ∈ ℝ⁸ derived from ARGIRA X: ε̂ₓ(i) = v̂ᵢ · εₓ where v̂ᵢ is the normalized feature vector associated with image i. This projection produces a continuous scalar ε̂ₓ for each image in a corpus of N = 227 images, enabling systematic evaluation of the relationship between texture-based descriptors and residual structural variation unexplained by the Tempo·Space decomposition. Analytical Structure The dataset supports four sequential analytical modules: XI-A0: Spatial collapse analysis between GLCM metrics and Tempo·Space representation XI-A1: Projection analysis and partial correlation modeling XI-A1b: Nested regression models with Leave-One-Out cross-validation and bootstrap validation XI-A0.5: Robustness and interpretability analysis of glcm_global_contrast Main Findings Most GLCM descriptors collapse onto the Tempo·Space manifold, indicating strong redundancy between texture features and the v3.5 sonification embedding. After conditioning on Tempo·Space, nearly all GLCM signal becomes non-significant except for glcm_global_contrast, which remains a stable independent predictor of ε̂ₓ. The best-performing model: ε̂ₓ ~ Tempo·Space + glcm_global_contrast yields: ΔR² ≈ 0.064 ΔLOO ≈ 0.056 Low multicollinearity (VIF < 1.1) Bootstrap confidence intervals excluding zero Remaining GLCM features fail out-of-sample validation, suggesting that residual texture structure is effectively low-dimensional under the given conditioning. The model explains approximately 33.6% of variance in ε̂ₓ, leaving substantial unexplained structure for future extensions (ARGIRA XII). Interpretation Scope This work does not claim a complete explanation of the residual structure identified in ARGIRA X. Instead, it identifies a stable, minimal explanatory direction within the GLCM feature space and constrains the effective dimensionality of texture-based predictors under Tempo·Space conditioning. Future work will extend analysis toward spectral, chromatic, gradient-based, and latent representations. Reproducibility All analyses are fully reproducible using standard Python scientific libraries: numpy, pandas, scipy, scikit-learn, statsmodels