Background: In the three-dimensional kinematic analysis of the trunk during human locomotion, a multi-segmental rigid-body model would be a better representation for the trunk compared with a single rigid-body model with regard to goodness-of-fit. However, there is a trade-off between data fitting and the simplicity of the model.
Research question: This study aimed to determine the optimal number of rigid-body segments during walking and running using Akaike's information criterion (AIC), which determines the model that has goodness-of-fit and is generalizable.
Methods: Empirically obtained kinematic data for the trunk during walking and running were fitted by one-, two-, three-, and six-linked rigid-body models using a nonlinear optimization algorithm. The relative quality of these models was assessed using their bias-corrected AIC (AICc) value.
Results: The AICc values of two- and three-linked rigid-body models were significantly smaller than those of one- or six-segment models for the walking trial. For the running trial, the AICc values of two-, three-, and six-segment models were significantly smaller than that of the single rigid-body model.
Discussion: These results suggest that both two- and three-linked rigid-body models would be better than the one- and six-linked rigid-body representations for analyzing trunk movement during walking, whereas the two-, three-, and six-linked models would be comparably well-balanced models in terms of both the goodness-of-fit and generalizability for running analysis.
Keywords: AIC; Generalizability; Goodness-of-fit; Modeling; Simplicity.
Copyright © 2020 The Authors. Published by Elsevier B.V. All rights reserved.