Descriptive statistics were used to describe the demographic
characteristics of all the respondents. A
bivariate-correlation analysis was then carried out to
evaluate any correlation between the three dimensions of
perceived causes of motorcycle accidents and the driving
violations carried out while riding a motorcycle. This was
followed by an independent t test to evaluate any significant
differences between motorcyclists who were involved
in accidents and motorcyclists who were accident free, in
terms of these dimensions and behaviours. For the prediction
of a motorcycle accident, univariate logistic regressions
were first performed to examine the effect of
different demographic variables, alignment with the
different dimensions of perceived accident causes, and the
extent of driving-violation behaviours. Variables reaching
p > .05 were then eliminated. The remaining variables
(that is, with p < .05) were then entered into a multivariate
backward stepwise logistic regression to determine their
contribution to predicting active involvement in accidents
amongst these motorcyclists. All statistical analyses were
performed using the IBM SPSS programme version 20.0 for
Windows (IBM SPSS Inc., Chicago, IL, USA); the significance
level was set at p < .05.