U. S. Department of Transportation
Download 2.66 Mb.
|
- Navigate this page:
- Controller Time on Shift
- Controller Age
- Relevant Training in the Last Year
- Controller Workload
Controller VariablesThese variables originate from the ATQA OE database and therefore only pertain to OE incidents. The variables in this section describe the controller or controller’s situation at the time of the incident. Employee Alerted to Incident By(ATQA OE) This variable indicates who alerted the controller to the incident. Recall that this is coded only for OE incidents; so in all cases the controller was at fault, though the incident may be first identified by a different party. The overall frequency of each response is presented in Figure 26. Table 120 and Table 121 present the distribution as well as the results of a Chi-Squared test. 
Figure – Frequency of Categories of Employee Alerted to Incident By Table – Observed Distribution of Employee Alerted to Incident By, by Severity
Table – Expected Distribution of Employee Alerted to Incident By, by Severity
The majority of incidents appear to be identified by persons other than the controller. Additionally, incidents identified by pilots tend to be more severe than expected. All categories except category D incidents are higher than expected (with category A being twice as high as expected). The opposite pattern holds for incidents identified by other facility personnel. The pattern is less clear for self-identified incidents, where categories A, B and D are lower than expected and category C is observed more frequently than expected. The deviations from the expected values are much higher for pilot identified incidents than for either self-identified or those identified by other personnel. Table 122 presents the results of a simple logit focusing on OE incidents identified by pilots. Table – Logit Estimate of Impact on Severity, Employee Alerted to Incident By, Conflict Only
The results indicate that the odds of an OE incident being severe if it is identified by a pilot are 3 times higher than incidents not identified by pilots. This is consistent with the information contained in Table 120. Future Research
One possible explanation for this pattern is that, due to their proximity, pilots are able to identify the most serious incidents. This would cause the increase in pilot-reported serious OE incidents. This trend may not be unique to OE incidents, but there is no counterpart variable describing PD incidents. Further research is warranted to better understand how severity and who identifies the incident are related. Controller Time on Shift(ATQA OE) This variable tracks the time the controller was on shift before the incident occurred. Again, this is only available for OE incidents. Figure 27 and Table 123 present the distribution of this variable while Table 124 presents the results of Kruskal-Wallis test by severity category. 
Figure – Distribution of Time on Shift Table – Percentiles of Time on Shift
Table – Kruskal-Wallis Test Results for Time on Shift
The overall distribution is confined mostly before 500 minutes. This is not entirely surprising, as shift length is regulated. However, it is worth noticing the observations above approximately 500 minutes. These observations are certainly outliers and may be misreported. However, the number is not large enough to distort the distribution and, without further information, the values are certainly possible if unlikely and so should not be excluded. Future Research
The distributions by severity level look fairly similar. This observation is borne out by the results of the Kruskal-Wallis test that indicate no joint difference between the groups. The most obvious explanation for this is that time on shift does not influence severity of the incident. It is possible that the frequency of incidents might go up as time on shift goes up.50 It is important to note that no information on controller shifts without incursions is available – the vast majority of shifts have no incursions. Further investigation into the relationship between time on shift and frequency of incursion is warranted. Controller Age(ATQA OE) This variable indicates the controller age in years. As this variable is derived from ATQA, it is only available for OE incidents. Table 125 and Figure 28 present the distribution of controller age while Table 126 gives the results of a Kruskal-Wallis test by severity. 
Figure – Distribution of Controller Age Table – Percentiles of Controller Age
Table – Kruskal-Wallis Test Results for Controller Age
There does not appear to be a relationship between controller age and incident severity. Controller age is a weak proxy for controller experience. A more focused look at controller experience may reveal a different pattern. Additionally, it is important to note that these results are in terms of severity and nothing can be said about the frequency with which controllers of a given age commit errors. Relevant Training in the Last Year(ATQA OE) This variable indicates whether the controller was involved in “relevant” training in the last year. Note that this is a self-reported variable on the controller incident reporting form. Additionally, no guidance is given on what constitutes relevant training. At a minimum it is assumed to be training broadly related to runway incursions. Table – Observed Distribution of Relevant Training in Last Year by Severity
Table – Expected Distribution of Relevant Training in Last Year by Severity
There does not appear to be any relationship between receiving training and severity. It is possible that training may affect the frequency with which errors occur, but no conclusion regarding frequency can be drawn from these results. Controller Workload(ATQA OE) Controller workload measures the number of aircraft the controller was responsible for at the time of the incident. This is a self-reported variable on the controller error reporting form. 
Figure – Distribution of Controller Workload Table – Percentiles of Controller Workload
Table – Kruskal-Wallis Test Results for Controller Workload
The test results indicate that the severity categories are jointly different in terms of controller workload. Further, all categories can be considered pairwise different from category D (no other pairwise comparisons are significantly different). Table 131 presents the results of a Kruskal-Wallis test for conflict events only. Once the conflict versus non-conflict dynamic has been eliminated, controller workload does not appear to have a different distribution by severity. Controller workload may serve as a proxy for the overall traffic level at an airport, rather than directly impacting severity. A more focused look at extreme controller workload levels may also reveal a different pattern (given that the overall distributions are fairly narrow). Table – Kruskal-Wallis Test Results for Controller Workload, Conflict Only
|