A Theoretical Analysis on the Impacts of Misestimated Test Cases

The outcomes of test play an important role in fault localization. Due to the oracle problem, there are a large number of test cases unlabelled in testing. They can not be used directly and it seriously affects the effectiveness of Spectrum-based fault localization techniques(SBFL). To utilize those test cases, various approaches have been proposed to estimate test outcomes. However, there is no guarantee that all unlabelled test cases are identified correctly. It is unknown that how these misestimated test cases will influence the performance of SBFL. In this paper, a theoretical analysis on the impacts of misestimated test cases is performed by considering two different types of misestimating. There are 24 risk evaluation formulas investigated in our study. Based on the analysis, some suggestions are provided for the labelling of unlabelled test cases to make it be beneficial to locate faults.