The statistics captured during testing a faulty program are the primary source of information for effective fault localization. A typical ranking metric estimates suspiciousness of executable statements and ranks them according to the estimated scores. The coverage-based ranking schemes, such as the metric used in Tarantula and Ochiai score, utilize the execution profile of each test case, including code coverage and the statistics associated with the number of failing and passing test cases. Although the coverage-based fault localization metrics could be extended to hypothesis testing and in particular to the chi-square test associated with crosstab or known as contingency tables, not all contingency table association metrics are explored and studied. We introduce the odds ratio metric and its application to the fault localization problem. The odds-ratio metric has been used extensively in categorical data analysis and in measuring the association of dependency between dichotomous variables. However, its application to fault localization metric is new. Furthermore, we investigate the effectiveness of conditional odds ratio metric for fault localization when there are multiple faults in the programs. Our experimental results show that the odds ratio metric performs better than the other ranking metrics studied for single faults, whereas, the conditional odds ratio ranking scheme is competitive when there are multiple faults in the software under test.