Data Analysis
Power was estimated to determine the number of participants
needed for each outcome measure (Short FES-I,
TUG, and the LTEQ) over a 6-week period. The overall
power was set to 0.80, and equal sample sizes were
assumed. A sample size needed of at least 45 participants
per group was determined using a multivariate method22
with a mean effect size of 0.25.
After calculating mean scores and times, a traditional
repeated-measures analysis of variance22 was completed
for each outcome measure (Short FES-I, TUG, and LTEQ).
Pre- and posttest change scores and times for the outcome
measures and their 95% confidence intervals were computed
to examine the intervention effects. A two-tailedalpha level was set to .05. Effect sizes were also interpreted
as a standardized measure of the size of the treatment
effect independent of the sample size. According to
previously used guidelines for partial-eta square (g2), g2 of
0.01 represents small effect sizes, g2 of 0.06 represents
medium effects, and g2 greater than 0.14 represents large
effects.23 All data analyses were performed using PASW
software version 18 (IBM, Chicago, IL).