Rethinking the Statistical Analysis of Neuromechanical Data

Abstract

Researchers in neuromechanics should upgrade their statistical toolbox. We propose linear mixed-effects models in place of commonly used statistical tests to better capture subject-specific baselines and treatment-associated effects that naturally occur in neuromechanics. Researchers can use this approach to handle sporadic missing data, avoid the assumption of conditional independence in observations, and successfully model complex experimental protocols.