In order to ensure the quality of data collected before starting
data analysis, all the completed questionnaires (N = 215) were
checked against systematic response patterns and more than 5%
missing items (Seo, 2005). Through this data screening process, 2
out of 215 completed questionnaires were dropped from data
set. The responses with 5% or less unanswered items were retained
for data analysis. Missing data points were imputed with the median
of nearby points in each case.
To code responses for data analysis, the researchers identified
each item as being favorable or unfavorable toward its factor to
be measured, as suggested by Seo (2005). For items that measure
management safety commitment, social support, safety motivation,
and safety knowledge, the higher the assigned value, the more
favorable these constructs were indicated. For items that measure
the production pressure, the higher the assigned value the higher
level of production pressure. For the safety behavior, the higher
values, the more frequent those safety behaviors were conducted.