The use of exponential smoothing to produce yearly updates of loss rate estimates in Marine Corps manpower models.
Hogan, Daniel L. Jr.
Read, Robert R.
Gafford, Jack B.
MetadataShow full item record
The use of exponential smoothing to perform yearly updating of attrition rates is examined and has merit. It shows enormous flexibility in adjusting to changes in the environment affecting the attrition rates, and it displays almost as much accuracy as the method it is intended to replace while using thousands of times less data. A secondary purpose of this study is fulfilled in confirming that the current aggregate methods are outperformed by maximum likelihood estimation, transform cell scale average, James-Stein, and limited translation James-Stein. None of these four methods is dominant overall, but all are improvements over the estimation system now employed.