Date of Conferral

2021

Degree

Ph.D.

School

Management

Advisor

Nikunja K. Swain

Abstract

Since the topic of improving data quality has not been addressed for the U.S. defense cost estimating discipline beyond changes in public policy, the goal of the study was to close this gap and provide empirical evidence that supports expanding options to improve software cost estimation data matrices for U.S. defense cost estimators. The purpose of this quantitative study was to test and measure the level of predictive accuracy of missing data theory techniques that were referenced as traditional approaches in the literature, compare each theories’ results to a complete data matrix used in support of the U.S. defense cost estimation discipline, and determine which theories rendered incomplete and missing data sets in a single data matrix most reliable and complete under eight missing value percentages. A quantitative pre-experimental research design, a one group pretest-posttest no control group design, empirically tested and measured the predictive accuracy of traditional missing data theory techniques typically used in non-cost estimating disciplines. The results from the pre-experiments on a representative U.S. defense software cost estimation data matrix obtained, a nonproprietary set of historical software effort, size, and schedule numerical data used at Defense Acquisition University revealed that single and multiple imputation techniques were two viable options to improve data quality since calculations fell within 20% of the original data value 16.4% and 18.6%, respectively. This study supports positive social change by investigating how cost estimators, engineering economists, and engineering managers could improve the reliability of their estimate forecasts, provide better estimate predictions, and ultimately reduce taxpayer funds that are spent to fund defense acquisition cost overruns.

Download

Share

 
COinS