Impact of variable transformations on multiple regression models for enhancing gait normalization

Abstract

Gait analysis has become an important tool in clinical practice for monitoring disease progression and evaluating therapeutic interventions. However, a subject's gait characteristics can be affected by physical characteristics such as age and height, which can interfere with accurate comparisons between subjects. MLR normalization has been shown to be effective in reducing interference from subject-specific physical properties, but non-linear effects can still impact the results. In this study, the independent variables were transformed to improve normalization performance, and the results indicate that using MR normalization with data transformation can effectively de-correlate physical characteristics from gait variables, improving the model fit and augment the capability to compare subjects with varying physical characteristics. This study provides valuable insights into the use of MLR models for gait normalization, with potential applications in clinical practice and research.