An Empirical Bayes Approach to Forecasting Marine Corps Enlisted Personnel Loss Rates
Abstract
This report describes an empirical Bayes approach to forecasting Marine Corps enlisted personnel loss rates. The context is a simple time series regression model. The report emphasizes a comparison of the accuracy of forecasts based on empirical Bayes estimates of parameters and the accuracy of forecasts based on least squares estimates of parameters. Additionally, an enlarged class of estimators is developed, the so-called double F empirical Bayes estimators. The test data consists of 24 quarterly observations (FY81 through FY86) on end-of-active service (EAS) loss rates for the entire enlisted Marine Corps and three representative occupational fields. For each series and each of parameter estimates, a mean square error of forecasts is computed to assess forecasting accuracy. Smaller mean square errors correspond to more accurate forecasts and larger mean square errors correspond to less accurate forecasts. All relevant accuracy measures are tabulated.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 01, 1988
- Accession Number
- ADA200236
Entities
People
- James P. Boyle
- Robert M. Holmes Jr.