The application of multilevel modelling to account for the influence of walking speed in gait analysis.

Differences in gait performance can be explained by variations in walking speed, which is a major analytical problem. Some investigators have standardised speed during testing, but this can result in an unnatural control of gait characteristics. Other investigators have developed test procedures where participants walking at their self-selected slow, preferred and fast speeds, with computation of gait characteristics at a standardised speed. However, this analysis is dependent upon an overlap in the ranges of gait speed observed within and between participants, and this is difficult to achieve under self-selected conditions. In this report a statistical analysis procedure is introduced that utilises multilevel modelling to analyse data from walking tests at self-selected speeds, without requiring an overlap in the range of speeds observed or the routine use of data transformations.