Statistical Techniques for Determining Officer Separation and Retirement Trends in the United States Air Force.
Abstract
The primary source of historical loss information is the Fiscal Year Computerized Officer Projection System report. The analysis techniques used were linear regression and Box and Jenkins' time series. The regression models developed for both separation and retirement predictions were very accurate. The FY81 separation prediction was in error by only 1.8% and the FY82 separation and retirement predictions were in error by 16.9% and 2.1% respectfully. Moreover, a modified update procedure was in error by only 9.6% for the FY82 separation prediction. This compares to errors in loss predictions of 1.7% to 79.9% for the currently used models. Although time domain time series models were developed which adequately fit both separation and retirement patterns, both failed to accurately predict either short or long term trends.
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 01, 1982
- Accession Number
- ADA124836
Entities
People
- Albert C. Dremstedt
Organizations
- Air Force Institute of Technology