In practice, this qualitative question is less interesting than the quantitative one of how much irregularity affected the results, and in particular whether irregularities altered the electoral outcome. Medzihorsky (2015) proposed a latent class to digit based election forensics. In this approach, the observed numerals are sorted into two classes: accurate and irregular. The classes are latent in the sense that their memberships are not observed and have to be inferred from the data.