Aggregation of Population Groups to Improve the Predictability of Marine Corps Officer Attrition Estimation

摘要

This thesis presents an algorithm for the aggregation of low inventory categories (small cells) which characterize the population of Marine Corps, unrestricted, active duty officers. The basis for aggregating these small cells is the degree of homogeneity of historical attrition rates. The techniques of hierarchical cluster analysis are applied to the small cell problem in lieu of existing functional and organizational structures. This research demonstrates the adaptability of cluster analysis to loss rate aggregation and provides a shell for more refined model applications. Further, statistical stability and attrition rate homogeneity have been introduced to allow for subsequent application of shrinkage type parameter estimation methods associated with the development of an officer attrition rate generator.