Date of Award


Document Type


Degree Name

Master of Science in Operations Research


Department of Operational Sciences

First Advisor

James T. Moore, PhD

Second Advisor

Jack M. Kloeber, PhD


This research has two objectives-to verify and validate the U.S. Army's Forecast and Allocation of Army Recruiting Resources (FAARR) model and to develop a Data Envelopment Analysis (DEA) modeling strategy. First, the FAARR model was verified using a simulation of a known production function and validated using sensitivity analysis and ex-post forecasts. FAARR model forecasts were not accurate and were extremely sensitive to any changes in the model's linear programming constraints and to changes in recruiting resource levels. Second, this research describes a three phase modeling strategy to build accurate DEA models. DEA has become a popular tool to evaluate the relative efficiency of many types of organizations. However, the literature has paid little attention to the practical problems of selecting the appropriate input variables and envelopment frontier. Analysts may use a number of diagnostic techniques to detect misspecification in statistics based models. No such diagnostics exist for DEA models. Without a-priori Knowledge concerning the production process's appropriate input variables and returns to scale, analysts do not know if they have constructed an accurate DEA model. Using a three phase strategy, relevant DEA model input variables are selected using Principal Component Analysis and Ordinary Least Squares (OLS) regression. The appropriate DEA envelopment frontier is selected using a Monte-Carlo simulation of an estimated production function representing the actual production process. The research concludes by demonstrating ex-post forecasts from a combined OLS/DEA model were more accurate when the DEA model selected by the three phase modeling strategy was used.

AFIT Designator


DTIC Accession Number