STATISTICALLY DEFENSIBLE TEST AND EVALUATION: THE AIR FORCE FLIGHT TEST CENTER PERSPECTIVE

摘要

Since 2004 the Air Force Flight Test Center (AFFTC) has been on a statistics journey -we have vigorously pursued an initiative to improve our use of statistics in planning tests and evaluating subsequent test results. Our initial efforts focused on design of experiment (DOE) techniques and we attempted to train all flight-test discipline engineers in the use of statistics in general and DOE in particular. However, after a few years we found that, although our engineers had a fair understanding of basic statistics, they often did not get to practice what they learned until months (or even years) after completing the statistics classes. This was due to many issues, such as program delays. Over time our approach to improve our test procedures and analysis has evolved to using a statistical consulting group and an associated broadening of the statistical techniques and methods. This paper is a brief summary of the AFFTC experience; it covers the technical and programmatic challenges encountered in implementation, including occasional push back from some of our technical experts, and examples describing the general use of statistics at AFFTC. The paper concludes with a brief discussion regarding future initiatives we are implementing to institutionalize the application of statistics to test and evaluation at AFFTC.