Most work has been on tests that focus on the vote counting process to detect patterns that are deemed unlikely to occur in genuine counts, but that might be a by-product of fabricating results. A second family of tests looks for surprising correlations between different aspects of the voting results. The range of fraud this second family of tests might uncover is more comprehensive. For example, they might detect systematic ballot invalidation during the vote counts in some areas, larger patterns of ballot-box stuffing by election officials, or even local coercion of voters to turn out and vote for one party. Thus, this second family might detect electoral malpractice that occurred on election day, not just in the tally. However, these tests rely on stronger assumptions about the mechanism that generates un-manipulated election results. In turn, that entails a higher risk of false alarms from patterns that arose from natural processes, rather than fraud. This report will discuss these limits below.