Table 2 shows the overall results per inspector and technique. To calculate the effectiveness indicator we have used 189, which is the total number of defects. We have performed a statistical analysis using the SPSS v 20.0.0 tool [20]. In this analysis, we used α = 0.05due to the small sample used within this study[3]. In order to compare the effectiveness and efficiency of both samples, we used the Mann-Whitney non parametrical statistic method and the boxplots graph to facilitate visualization.
Figure 4 shows the boxplots graph for the effectiveness and efficiency indicators. The results from the statistical analysis support both null hypotheses H01 and H02 indicating that there is no significant difference in the effectiveness indicator when using the DUE technologies with or without training (p = 0,481); and that there is no significant difference in the efficiency indicator when using the DUE technologies with or without training (p = 0,139). Moreover, the graph in Figure 4 shows that, for effectiveness, the median is similar in both treatments, which suggests that perhaps, the documentation for using the technologies is sufficient to enable their use by the software development industry. However, we can argue that given the small sample used, it is difficult to obtain statistical significance. Thus, the results should be considered indicators and not conclusive. In fact, Figure 4 shows that the group that received training obtained slightly higher results in the efficiencyindicator. We will explain the difficulty met by the subjects who did not receive training in the next subsection.