I am currently engaged in research that compares the performance of two correction tools across various error types. In my dataset, I've employed 2x2 contingency tables to categorize the frequency of error detection, with rows representing the tools (Tool A vs. Tool B) and columns indicating errors detected correctly and errors detected but actually not errors. For some error types, one tool failed to detect any errors, resulting in two cells with zero counts in the contingency tables. This issue has presented a challenge in calculating odds ratios or risk ratios and conducting chi-square or Fisher's exact tests. So is there any method to solve this problem?

Thank you very much for taking the time to assist me.