There are a number of approaches to take into account potential outside influences, and thereby potentially to improve the validity of statistical procedures used in election forensics— but none is perfect. Hicken and Mebane (2015) provide a package of statistical tests for anomalies in election results. They do not merely apply multiple tests in parallel; they analyse whether several indicators of irregularities agree, arguing that multiple indicators strengthen the evidence. However, particular attention must be paid to the nature of alternative processes that might produce similar anomalies; e.g., variance in the strategic context of an election between districts might affect many tests, and affect both the partisan distribution of the results and the turnout. Furthermore, because some irregularities in the election process are only identifiable by one test, they will not be corroborated by multiple tests. And, as mentioned previously, the null hypotheses tested involve rather stylized, counterfactual assumptions about how accurate election results would be generated. To reject the stylized model is not equivalent to showing that there was an irregularity. It merely means that the data would be surprising if the stylized model were true.