FEDERAL AVIATION ADMINISTRATION
Figure 1a shows an example of one category of data we frequently see in general aviation (GA) accident analysis. This is a histogram of fatal accident counts for instrument-rated pilots of GA aircraft1 as a function of pilots’ total flight hours (TFH).2 The data reflect actual U.S.NationalTransportationSafetyBoardaccidentdata from 1983-2000, inclusive (NTSB, 2011).
It is not hard to appreciate the usefulness of a modeling function here. Such a function would smooth the noise in the data, allowing investigators to better predict how many pilots of a given experience level are likely to be involved in accidents over a given time period. This would be useful, for instance, in allocating resources for pilot training, or as the basis for a statistical covariate of flight risk. Even a casual glance at Figure 1 shows that policy makers would want to focus on pilots having fewer than 5000 TFH, simply because there are far more accidents in that range. The question is how to get beyond the considerable noise in the data to arrive at more precise estimates of this kind.
About the FAA
“Our continuing mission is to provide the safest, most efficient aerospace system in the world…We strive to reach the next level of safety, efficiency, environmental responsibility and global leadership. We are accountable to the American public and our stakeholders.”