Adaptive Forecasting of the Size of a Force Subject to Random Withdrawals.
Abstract
The study develops a new adaptive procedure for the forecasting of the total size of a military organization (the Marine Corps), subjected to random withdrawals of first enlistees and of career force personnel, with unknown attrition rates. The attrition rates of various subpopulations are not the same. The total population of first enlistees are classified to homogeneous subpopulations according to the following factors: length of contract, period of entrance to service, education and race. Based on cohorts having the same levels of these factors prediction intervals (with 95% confidence level) for the future size of these cohorts are estimated. Three methods of estimating prediction intervals are compared: conditional maximum likelihood, tolerance intervals and Bayes prediction intervals. Analysis of actual Marine Corps data illustrates and effectiveness of the forecasting procedures. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 15, 1976
- Accession Number
- ADA031477
Entities
People
- S. E. Haber
- Shelemyahu Zacks
Organizations
- Case Western Reserve University