Support minimum sample size for uniform sampling, to allow more flexibility and restore clarity for opportunistic auditing #2183
Description
Support for Card Style Data ballot sampling (#2150) is fantastic, thank you!
But using this option greatly complicates opportunistic auditing, since the sampling will be much more likely to be non-uniform for all the contests which are not targeted, making it often impossible to calculate a risk level.
Also, years of experience in Colorado shows that even with uniform sampling, it is difficult to collect good samples for the opportunistic contests, since the only way to drive the sample size is by selecting a contest to target with just the right diluted margin and risk limit, as described at Colorado's Risk-Limiting Tabulation Audits: Accomplishments and Proposed Improvements.
A good solution would be to allow for a minimum sample size for each jurisdiction (each CVR file), and to choose that sample uniformly at random. Many outside observers also would like to see a larger uniform sample size for QA or to make the process easier to explain and justify to those who don't understand the statistics.