tri.calc.sample {elec} | R Documentation |
Calculate an estimated sample size to do a trinomial bound that would have a specified power (the chance to certify assuming a given estimate of low-error error rate), and a specified maximum risk of erroneously certifying if the actual election outcome is wrong.
tri.calc.sample(Z, beta = 0.75, guess.N = 20, p_d = 0.1, swing = 5, power = 0.9, bound = c("e.plus", "WPM", "passed"))
Z |
elec.data object |
beta |
1-beta is the acceptable risk of failing to notice that a full manual count is needed given an election with an actual outcome different from the semi-official outcome. |
guess.N |
The guessed needed sample size. |
p_d |
For the alternate: estimate of the proportion of precincts that have error. |
swing |
For the alternate: estimate of the max size of an error in votes, given that error exists. |
power |
The desired power of the test against the specified alternate defined by p_d and swing. |
bound |
e.plus, WPM, or use the passed, previously computed, e.max values in the Z object. |
An audit.plan.tri
object. This is an object that holds information on
how many samples are needed in the audit, the maximum amount of
potential overstatement in the election, and a few other things.
Luke W. Miratrix
See Luke W. Miratrix and Philip B. Stark. (2009) Election Audits using a Trinomial Bound. http://www.stat.berkeley.edu/~stark
See elec.data
for information on the object that holds
vote counts. See
tri.sample
for drawing the actual sample. See
audit.plan.tri
for theo object that holds the audit plan
information (e.g., number of draws, estimated work in ballots to
audit, etc.). See trinomial.bound
for analyzing the
data once the audit results are in. See tri.audit.sim
for simulating audits using this method. See CAST for an SRS audit method.
data(santa.cruz) Z = elec.data( santa.cruz, C.names=c("danner","leopold") ) tri.calc.sample( Z, beta=0.75, guess.N = 10, p_d = 0.05, swing=10, power=0.9, bound="e.plus" )