Date of Award
3-23-2017
Document Type
Thesis
Degree Name
Master of Science in Cost Analysis
Department
Department of Mathematics and Statistics
First Advisor
Edward D. White, PhD.
Abstract
The main concern of a program manager is to manage the cost, schedule, and performance triad of a program. Historically, programs tend to meet the performance aspect at the expense of cost or schedule, or both. This research gives the acquisition community a set of tools that enables them to impartially analyze the cost and schedule of their programs, helping to mitigate these issues. Five regression models encompass this toolset; one to estimate the median program cost and four to identify the probability of realizing a given overrun. The cost model explains 81% of the variation in program acquisition using seven predictor variables available to the estimator at the time of MS B start. Four logistic models estimate the probabilities that a program may identify as a program that experiences cost and schedule overruns of specific magnitudes from their MS B estimate. These models predict the group the program may reside in with an accuracy of at least 0.79 probability and use multiple predictor variables available at MS B. With these tools the program manager has the ability to preemptively identify potential problems in their program based on the program’s characteristics, potentially saving millions in cost and schedule overruns.
AFIT Designator
AFIT-ENC-MS-17-M-231
DTIC Accession Number
AD1051598
Recommended Citation
Trudelle, Ryan C., "Using Multiple and Logistic Regression to Estimate the Median Will-Cost and Probability of Cost and Schedule Overrun for Program Managers" (2017). Theses and Dissertations. 784.
https://scholar.afit.edu/etd/784