Logistic Regression Analysis of Operational Errors and Routine Operations Using Sector Characteristics

摘要

Two separate logistic regression analyses were conducted for low- and high-altitude sectors to determine whether a set of dynamic sector characteristics variables could reliably discriminate between operational error (OE) and routine operation (RO) traffic samples. OE data were derived from SATORI re-creations of OEs occurring at the Indianapolis Air Route Traffic Control Center between 9/17/2001 and 12/10/2003. RO data were extracted from System Analysis Recordings (SARs) taped between 5/8/2003 and 5/10/2003. Dynamic sector characteristics submitted as potential predictors were: Average Control Duration, Number of Handoffs, Number of Heading Changes, Number of Intersecting Flight Paths, Number of Point Outs, and Number of Transitioning Aircraft. In the low-altitude sector model, backward stepwise elimination reduced the variables to the Number of Intersecting Flight Paths, the Number of Point Outs, and the Number of Handoffs with 75% overall classification accuracy. In the high-altitude sector model, backward stepwise elimination reduced the variables to the Number of Intersecting Flight Paths, the Number of Heading Changes, the Number of Transitioning Aircraft, and Average Control Duration with 79% overall classification accuracy. Classification rates achieved through the use of the selected sector characteristics support the assumption that elements of the sector environment contribute to the occurrence of OEs. Continued investigations along these lines may highlight complexity factors that should be addressed to ensure that separation is maintained.