Date of Award
Master of Science in Cost Analysis
Daniel V. Ferens, PhD
As more Department of Defense resources are being allocated toward software development, the necessity to accurately plan for software costs has become critical. Obtaining reliable estimates from software cost models, like PRICE S, can be a problem when input parameters are not precisely defined or calibrated. This research effort centered on refining Productivity Factor (PROFAC) values for defense industry applications. The Space and Missile Systems Center Database was used to calibrate PROFAC values for eleven stratified data sets: military ground, military mobile, missile, unmanned space, Ada, Assembly, C, COBOL, FORTRAN, JOVIAL, and PASCAL. The accuracy of the calibrations was determined through comparisons of calibrated and default generated estimates versus actuals. Statistical methods used to make the comparisons included standard deviation, mean absolute error, mean relative error, and percentage of records estimated within twenty-five percent of actuals. The results were surprising in that, in most instances, the calibrated PROFAC values estimated actual cost well, but not overwhelmingly better than the default PROFAC values. The main contributing factor to this phenomena was variability within the stratified data sets. The results were encouraging, however, in that the results from seven of the eleven stratified data sets suggested either a new refinement in PROFAC values based upon the calibration or the recommendation to use PROFAC values from analogous calibrated records for estimating future efforts.
DTIC Accession Number
Galonsky, James C., "Calibration of the Price S Software Cost Model" (1995). Theses and Dissertations. 6505.