Reproducibility package for "Observation-Conditioned Recoverability of an H2AT Stabilizing Input: A Computational Biomechanics Validity Analysis" - V2.1

This repository provides the reproducibility materials supporting the manuscript “Observation-Conditioned Recoverability of an H2AT Stabilizing Input: A Computational Biomechanics Validity Analysis.” The study investigates a distinction between exact model-level differential reconstructibility and observation-conditioned recoverability of the stabilizing input in the nonlinear Head-Two-Arms-Trunk (H2AT) sitting model. The computational analysis examines how sampling and differentiation scale, measurement uncertainty, temporal content, derivative representation, noise structure, and selected biomechanical parameter mismatch condition the numerical recovery of an exactly reconstructible model input. The repository contains the computational materials underlying analyses B0–B12, including analytical uncertainty validation, Sobol-based state-bound screening, differentiation-bias analysis, temporal-content studies, representation and noise-model robustness tests, parameter-mismatch analysis, common-support and search-bound audits, initial-condition and non-sinusoidal robustness studies, and trajectory-level Monte Carlo qualification of the retrospective reference-recovery metric. It also includes the final numerical solver qualification and finite-Monte-Carlo objective-gap analysis used to assess the robustness of the reported computational conclusions. The package includes executable analysis scripts, numerical inputs and outputs, tabulated results, qualification records, figures, provenance information, computational-environment specifications, checksums, and the accompanying supplementary material. The evidence and quantitative conclusions are intentionally limited to the specified H2AT model and tested computational conditions. The repository does not contain patient or clinical data, and the study does not claim experimental validation, physiological validity, clinical accuracy, estimator-independent performance, or quantitative transfer to other biomechanical models.