Craig (2001) hypothesized a 'killing zone'--a range of pilot total flight hours (TFH) from about 50-350, over which general aviation (GA) pilots are at greatest risk. The current work tested a number of candidate modeling functions on eight samples of National Transportation Safety Board GA accident data encompassing the years 1983-2011. The goal was largely atheoretical, being merely to show that such data can be modeled. While log-normal and Weibull probability density functions (pdf) appeared capable of fitting these data, there was some pragmatic advantage to using a gamma pdf. A gamma pdf allows estimation of confidence intervals around the fitting function itself. Log- transformation of TFH proved critical to the success of these data-fits. Untransformed TFH frequently led to catastrophic fit-failure. Due to the nature of the data, it may be advisable to place the greatest prediction confidence in a middle range of TFH, perhaps from 50-5,000. Fortunately, that is also the range that captures the vast majority of all GA pilots. With some care, GA accident frequencies appear predictable from TFH, given data parsed by (a) pilot instrument rating and (b) seriousness of accident. Goodness-of-fit (R2) tended to be excellent for non-instrument-rated pilot data and good for instrument- rated data. Estimates of median TFH were derived for each dataset, which will be useful to aviation policy makers. These data suggest that the killing zone proposed by Craig may be wider than originally believed.
Predicting General Aviation Accident Frequency From Pilot Total Flight Hours
2012
28 pages
Report
No indication
English
Air Transportation , Statistical Analysis , Aviation accidents , Pilots , Weibull density functions , Confidence level , Fitting functions(Mathematics) , Goodness of fit tests , Mathematical prediction , Poisson ratio , Probability density functions , Risk , Tfh(Total flight hours) , Ga(General aviation) , Modeling , Pdf(Probability density functions)