Cross-dataset generalization of kinematics-based lower-limb joint-moment prediction

This reproducibility package supports the manuscript “Cross-dataset generalization of kinematics-based lower-limb joint-moment prediction.” Ridge regression, a pointwise multilayer perceptron (MLP), and a temporal convolutional network (TCN) were developed using 224 Camargo gait cycles from 22 participants and evaluated internally using participant-disjoint out-of-fold predictions. The frozen models were externally evaluated on 1,276 Scherpereel cycles from 12 participants; the primary confirmatory Tier A cohort comprised 1,127 cycles from 10 participants. The archive contains de-identified processed arrays, trained MLP and TCN models, frozen Ridge, MLP, and TCN predictions, source-to-analysis MATLAB and Python programs, a 1,363,500-row prediction master table, 5,000-replicate participant-cluster bootstrap outputs, audit evidence, manuscript tables, and publication figures. Raw source datasets are not redistributed. The official source repositories and attribution requirements are documented in the archive. Derived data, results, documentation, tables, and figures are licensed under the Creative Commons Attribution 4.0 International License. Original software code is licensed under the MIT License. Corresponding author: Shun Jiang ([e-mail]). Funding: This research received no specific grant from funding agencies in the public, commercial, or not-for-profit sectors. Competing interests: The authors declare no known competing financial interests or personal relationships that could have influenced the work. Acknowledgments: The authors thank the teams of Camargo et al. (2021) and Scherpereel et al. (2023) for creating and publicly sharing the lower-limb biomechanics datasets used in this secondary analysis, and thank the participants whose contributions made the original datasets possible.