Probability of Success in Primary Flight Training as a Function of ASTB Scores and API Grades: An Example of the Statistical Inferencing Component of the Pilot Prediction System

摘要

The Pilot Prediction System (PPS) is a research effort designed to provide Navy managers and other decision makers with improved access to selection and training data. Many disparate data bases, each containing partial and sometimes overlapping information on selection data and training performance, currently exist. There has been no attempt to coordinate the bits and pieces gathered into these local databases into a coherent whole. Such data needs to be merged and the anomalies excised so that a more global picture of selection data and training performance can emerge. In addition, the targeted users of the PPS should be shielded from the low-level technicalities of the data base because such technical details are of no concern to them. For the same reason, the statistical manipulations that provide extrapolations from the data base to new cases can be hidden from view. This report documents the first efforts at constructing the statistical modeling component of the PPS as derived from Bayesian statistical decision theory. It enables the potential user of the PPS to predict success in primary flight training for flight students based on their scores on selection tests and ground school performance.