On quality of test sets: relating fault coverage to defect coverage

The true quality of a test set can be evaluated by estimating its coverage of arbitrary defects. In evaluating the quality of test sets, test vectors were generated from different test generators. We estimate the corresponding defect coverages for these test sets of varying lengths, generated by different test generators. The coverage estimation is performed using a region based model, in which it is independent of specific, physical, fault models. Experimental results show that similar fault coverages of varying test set sizes yield similar defect coverage. Further, similar activity and observability levels were observed in spite of the fact that test sets were generated with different methods and had different sizes