The work attendance data were pooled to enable the percentage of maintenance
personnel present at work at each hour of the day to be estimated
(Figure 1). It can be seen, for example, that between 35 and 40% of
respondents indicated that they were at work between 01:00 and 05:00 h
during their most recently worked shift, and over 50% of workers
reported being at work between 08:00 and 16:00 h. A correction factor
was calculated to express the difference between the estimated percentage
of personnel present at work each hour and a nominal value of 50% work
attendance. Error frequencies were then multiplied by the correction
factor to arrive at the number of errors expected if an equal number of
workers had been present at work each hour of the day.
Figure 2 (top panel) presents the distribution of skill-based errors,
rule-based mistakes, knowledge-based mistakes, and procedure violations
over the 24 h before correcting for work attendance patterns. Figure 2
(bottom panel) shows the frequencies after correction for work attendance
patterns. It can be seen that skill-based errors were most frequent in the
early hours of the morning, diminishing in frequency throughout the day
until just after midnight, from which time they began to become more
prevalent. These patterns were apparent in both the corrected and uncorrected
data. Rule-based mistakes, knowledge-based mistakes, and procedure
violations exhibited no such clear pattern throughout the 24 h.
One-sample Kolmogorov-Smirnov tests were conducted to test the
null hypothesis that the corrected hourly frequencies for each error type,