Extension of Aggregation and Shrinkage Techniques Used in the Estimation of Marine Corps Officer Attrition Rates
Abstract
In this thesis we treat the 'small cell' problem encountered when building an attrition rate generator for large-scale manpower flow models. specifically for the USMC Officer Corps. Such models have a large number of low- inventory (i.e. small) personnel cells. This presents a dilemma: on one hand we want to preserve as much fidelity as possible in our work by preserving a great deal of detail in each cell; on the other hand our statistical estimation techniques require larger cell sample sizes than intrinsically occur cell-by- cell in actual sample data. Our approach to producing stable attrition rates for such cells involves two efforts: (i) the aggregation of cells into groups that exhibit homogeneity of attrition behavior, and (ii) and the development of 'shrinkage' estimation techniques for use in the individual groups. Aheuristic algorithm is developed and tested to treat the aggregation problem. Empirical Bayes methods are developed to serve the multi-cell estimation requirements needed to preserve the fidelity. Cross validation techniques are used to verify these methods. The present work builds upon the results of previous studies; we integrate what was learned into a coherent package that is ready for use. Keywords: Personnel attrition; Personnel management; Marine corps officers; Aggregation; Attrition rate estimation; Empirical Bayes.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 01, 1989
- Accession Number
- ADA219266
Entities
People
- John M. Misiewicz
Organizations
- Naval Postgraduate School