2.3. Statistical analysis
The statistical analysis of the data obtained from processing the
questionnaires was performed using IBM SPSS (Version 21.0.
Armonk, NY: IBM Corp). In order to verify the soundness of the
basic hypothesis in parametric testing, the Levene test was used to
check the homogenous variance assumption and the ShapiroeWilk
test to verify the normality assumption. The normality hypothesis
was invalidated (p < 0.001) for both knowledge and practices. For
one grouping variable, i.e. professional experience, the Levene test
rejected the hypothesis of equal variances (p ¼ 0.007) in the case of
practices. By consequence the omnibus tests for group differences
were realized using Kuskal-Wallis/ManneWhitney non-parametric
test. The data set being unbalanced, with the number of observations
(Na) as low as three for the age group of twenty and below
(Table 1), the exact significance, based on Monte Carlo simulation
(10,000 samples), was considered. Because the overall difference
test was found significant for one grouping variable with ordered
levels, i.e. education, the JonckheereeTerpstra multiple comparisons
procedure, with Bonferroni correction, was applied in this
particular situation. Variables found significant in the preliminary
univariate analysis were further investigated using the General
Linear Model (GLM) on ranked data. The interaction effect was
tested using the aligned rank method. Spearman rank correlation
coefficient was used to see how well knowledge and practice are
correlated. The statistical significance was set at the 0.05 level.