elec-package {elec}

Statistical Election Audits Package

Description

This is a collection of functions written to do various sorts of statistical election audits. There are also functions to generate simulated voting data, and simulated “truth” so as to do simulations to check charactaristics of these methods. The package includes two data sets consisting of actual reported voting results for races held November, 2008, in California. It also includes actual audit date for one of these races.

Package:	elec
Type:	Package
Version:	0.1
Date:	2009-01-14
License:	GPL (>= 2)
LazyLoad:	yes

There are three general audit styles implemented in this package. For each style there are two main computational tasks provided: estimate the needed sample size and expected workload, and calculate $P$-values for a given audit result. The three methods are CAST (see CAST.calc.sample and CAST.audit, the Trinomial Bound (see tri.calc.sample and trinomial.audit), and the Kaplan-Markov (KM) Bound (see KM.calc.sample and KM.audit).

The examples primarily use a data set included in the package, santa.cruz and santa.cruz.audit, which holds the ballot counts for a Santa Cruz, CA race that we audited using these methods. See trinomial.bound for how these data were analyzed. The yolo data set holds precinct level counts for a race in Yolo county.

There are also many functions allowing for construction of new audit methods and simulations. This includes methods that generate fake race data that can be used for computational simulations to assess the efficay of different auditing approaches (see, e.g., make.sample and make.truth).

The package grew out of an earlier, disorganized package that implemented general routines for election auditing. Pieces of this package are used by the aforementioned cleaner methods, but all the individual functions are still there for specific uses, such as making different tests. Start with stark.test, which has an index of these pieces in its “see also” section.

If you find yourself confused, please contact the maintainer, L. Miratrix, for help. This will help improve the clarity of the package a great deal.

Author(s)

Luke W. Miratrix

Maintainer: Luke W. Miratrix <[e-mail]>

References

CAST and KM were developed by Philip B. Stark. The Trinomial bound was developed by Luke W. Miratrix and Philip B. Stark.

For general papers on election auditing see the list at http://www.stat.berkeley.edu/~stark/Vote/index.htm.

In particular, for the trinomial bound, see Luke W. Miratrix and Philip B. Stark. (2009) Election Audits using a Trinomial Bound (in press).

For the KM bound see Stark, P.B., 2009. Risk-limiting post-election audits: P-values from common probability inequalities.

For an overview of the races and the methods, see Joseph Lorenzo Hall, Philip B. Stark, Luke W. Miratrix, Elaine Ginnold, Freddie Oakley, Tom Stanionis, and Gail Pellerin. (2009) Implementing Risk-Limiting Audits in California.