This program calculates the audit sample size, S, based on the estimate in Aslam, Popa & Rivest's
"On Estimating the Size and Confidence of a Statistical Audit"1 and the parameters in Stanislevic's
"Random Auditing of E-Voting Systems: How Much is Enough?"2

1. Enter name of this race (e.g., "US House"):

2. Enter total precincts, machines or systems in this race (e.g., "400"):

3. Enter total votes cast in this race (e.g., "100000"):

4. Enter reported margin in votes (e.g., "10000"):

5. Enter desired confidence level as a % (e.g., "99"):

6. Enter max. within-precinct miscount as a % of total precinct vote (WPM, e.g., "20"):3

7. Enter min. number of precincts, machines or systems that could change the outcome (e.g., "60"):4
(Note: If this data is not yet available, leave this field blank. The program will estimate a sample size based on the other available parameters.)

References:
 1. Aslam, Popa and Rivest, On Estimating the Size and Confidence of a Statistical Audit
2. Stanislevic, Random Auditing of E-Voting Systems: How Much is Enough?
3. Saltman referred to this as the maximum level of undetectability by observation in his 1975 work, Effective Use of Computing Technology in Vote-Tallying.
4. Required for precinct size adjustment. See Stanislevic, Pg. 15, Step 4. Sort all precincts, machines or systems in this race by the total votes cast in this race. Count the number of precincts, machines or systems (starting with the one with the largest vote count) needed to equal or exceed the value of (Margin / WPM / 2). This gives us the minimum number of corrupt precincts, machines or systems that could reverse the outcome of the race. If this data is not currently available, the estimated number displayed in Item (F) on the next page can be used to start the audit immediately. The final determination of the audit sample size should only be made using the actual precinct/machine/system-level vote count rankings.