Application of logistic regression and survival analysis to the study of CEP, manpower performance and attrition

Abstract

This thesis is an application of logistic regression and survival analysis techniques to the study of current estimated potential (CEP), manpower performance, and attrition behavior in the Singapore military. The manpower data includes both active (30%) and reserve personnel (70%) who entered service as early as the late fifties to as recent as the year 1992. The covariates under consideration are education level, academic or overseas military training award, current rank, length of service, rank seniority, age, salary grade, previous year's annual performance grade and CEP estimates. The study identifies the covariates that explain the CEP and annual performance for the binary and polytomous models of the officers who were still on active duty as of 31 Dec 1992. It also examines the trend of attrition behavior of officers using data from both the active and reserve personnel. The results of the study show that (1) higher education level does not necessary result in better performance grade although it seems to give an indication of higher CEP, (2) The higher the rank of an officer, the more likely it is for him to have a poorer performance grade than when he was in the previous rank, (3) Education level is a significant covariate of the survival functions, and (4) Engineering officers generally has a higher attrition rate than the other service support officers.