Defense Acquisition Research Journal Issue 95

January 2021

prime contractor or subcontractor (shown in the Contractor Designation column of Table 3). One of the most widely used hypothesis test techniques is a parametric test, such as the t -test. However, an underlying assumption of parametric tests is that the data are normally distributed. Therefore, a Shapiro-Wilk test was conducted to determine whether or not the data were normally distributed. The results of the test showed that the data were not normally distributed, thereby indicating parametric techniques should not be used. As a result, nonparametric tests (which do not require the assumption of normality) are utilized throughout the remainder of the analysis. Specific nonparametric tests used include the Kruskal-Wallis and Steel-Dwass tests, which are similar to ANOVA and t -tests. The Kruskal-Wallis test is a rank-based nonparametric test to determine whether statistically sig nificant differences exist between two or more groups of an independent variable on a continuous dependent variable. The dependent variable is the numerical cost factor value, while the independent variables are the various groups. For example, contractor type (prime versus subcontractor) is the independent variable, while the cost factor values are the dependent variable. Because the Kruskal-Wallis test does not identify where within the subcategory comparison differences occur, the Steel-Dwass test is employed. The Steel-Dwass multiple comparison test identifies which rank orders of the tested groups are statistically different for each instance of The Kruskal-Wallis test is a rank-based nonparametric test to determine whether statistically significant differences exist between two or more groups of an independent variable on a continuous dependent variable.

49

Defense ARJ, January 2021, Vol. 28 No. 1 : 40-70