Article

When analysing the reliability of ratio-scaled variables, such as walking
energy cost, variability of the error term often increases with increasing mean
values. This phenomenon is called heteroscedasticity, and it makes the analysis
of reliability more complicated. This study presents an examination of
heteroscedasticity for walking energy cost before analysing the reliability.
METHOD: Walking energy cost was collected from 33 children with cerebral palsy
(CP), with varying Gross Motor Function Classification System (GMFCS) levels (19
males; 14 females; mean age: 7y 6mo [SD 2y 6mo]; GMFCS levels I [n=16], II [n=7],
and III [n=10]). It was assessed by measuring oxygen uptake during 10 minutes of
resting and 5 minutes of walking at comfortable speed. Measurements were
performed twice, within 4 to 6 weeks. Primary outcomes included gross energy
cost, gross non-dimensional energy cost, net energy cost, net non-dimensional
energy cost, speed, and non-dimensional speed. Heteroscedasticity was analysed
with Bland-Altman plots and Kendall's tau. RESULTS: Visual inspection of the
Bland-Altman plots showed heteroscedasticity for gross energy cost, gross
non-energy cost, and net energy cost. This was confirmed by Kendall's tau
coefficients. Accordingly, data were logarithmically transformed, and reliability
was assessed with ratio statistics. For speed, heteroscedasticity was not
observed. INTERPRETATION: Variability of gross energy cost, gross non-energy
cost, and net energy cost, assessed across different GMFCS levels in children
with CP, was proportional to the mean, indicating the presence of
heteroscedasticity. This finding emphasizes the importance of always performing a
heteroscedasticity examination in reliability studies on energy cost and
reporting the reliability statistics accordingly.
CI - (c) The Authors. Developmental Medicine & Child Neurology (c) 2011 Mac Keith
Press.

Langue : ANGLAIS