~ tracking down this book ~
It is shown that the use of regression methods in the forecasting of Separations (EAOS), Eligibles (to reenlist) and Non-reenlistments jointly by length of service and pay grade are competitive with the currently used 'alpha' method. The question of whether one of the two methods of forecasting is clearly superior could not be addressed with the currently available data. The report describes the data base, presents various general characteristics of the data, summarizes the computational results that lead to the recommended choice of input, and recommends follow-on work to clarify the issues. (Author)